International Journal of Intelligent Systems and Applications in Engineering

Forecasting Air Quality with Deep Learning

2024-06-11T09:38:42+00:00

Due to factors such as increased urbanization, growing populations, transportation, home activities, agricultural methods, and industrial processes, Air pollution has emerged as a major issue in the past several years. It is linked to several illnesses and has emerged as a substantial issue in several urban areas, particularly in developing nations such as India. As part of our research, we make use of the Air Quality Index (AQI) for assess the quality of the air in Mumbai, India. Our emphasis is on evaluating 13 different pollutants and 7 meteorological indicators for the period from July 2017 to September 2022 to be able to predict air pollution levels. We employed three deep learning models: LSTM, Bi-LSTM, and CNN-Bi-LSTM. Results show that the CNN-Bi-LSTM model had better accuracy compared to earlier models, as proven by a MAE of 0.45, a MSE of 0.58, a RMSE of 0.60, and RMSLE of 0.36. This study shows that deep learning model are effective in forecasting AQI and by using historic data and deep learning algorithms enable precise forecasts of urban air quality levels worldwide.

Optimizing Resource Allocation in Cloud Systems using Reinforcement Learning Driven Dynamic VM Placement

2024-06-11T09:41:07+00:00

Virtual machines (VMs) are extensively used these days as a substitute for physical machines. When the computation power requirement goes beyond that of the existing physical systems, based on client-specific memory requirements, their tools subscriptions, and services the appropriate number of VMs needs to be allocated dynamically. The aim is to minimize the resource cost and energy consumption for optimal usage and enhancement of savings. This is hence an optimization problem that needs to be addressed based on various parameters linked to the system. In this paper, we have worked towards the allocation or placement of VMs in a cloud system, where based on previous requirements we train a model by reinforcement using the A3C algorithm, considering the replays of experiences in various states of the environment to ensure optimal allocation of VMs and hence the real-time functionality of the cloud system.

Lung Cancer Detection, Prediction and Analysis of Lifestyle Parameters using ML and AI Techniques

2024-06-11T09:43:10+00:00

Cancer poses a significant threat to human life, often diagnosed in later stages, highlighting the crucial need for early prediction. Literature extensively explores Machine Learning, Data Mining, and Artificial Intelligence techniques for the identification, classification, prediction, and detection of various cancers like lung, breast, prostate, skin, liver, and recurrence cancer. Predictive models rely on vast datasets for cancer prediction. Lung cancer's development is closely tied to lifestyle factors such as smoking, air pollution, and diet imbalance, emphasizing the potential of lifestyle indicators in early detection. A study focuses on constructing a model using lifestyle data to predict lung cancer and categorize its severity. Basic lifestyle parameters are initially examined, and if lung cancer potential is indicated, a second component of the model further analyzes each parameter to predict cancer level. Various Machine Learning techniques including Support Vector Machine, Logistic Regression, and Linear regression are applied to predict lung cancer risk and level with analysis of prediction using AI techniques. SVM emerges as the most effective classifier for predicting lung cancer risk based on lifestyle factors, while linear regression is optimal for total risk score prediction. Additionally, gender- and age-specific lifestyle parameters contributing to lung cancer are identified. The study's preliminary phase employs logistic regression and Support Vector Machine to predict lung cancer, achieving high accuracies of 94% and 90% respectively. The subsequent component utilizes SVM, Random Forest, KNN and Linear Regression to estimate cancer malignancy levels, with accuracies reaching 98% , 96 and 97% respectively. The study aims to predict lung cancer early using lifestyle data, offering insights into risk factors and preventive strategies.

Garbage Classification based on Dense Network (GCDN) using Transfer Learning and Modified Hyper Parameter

2024-06-11T09:45:13+00:00

Garbage classification plays a vital role in waste management and sustainability of the environment. Traditional methods of waste classification often depend on manual sorting, which is very time-consuming and prone to human errors which can lead to policy inadequacy by the government. In this paper, we proposed a deep learning-based DENSNET201 approach Garbage Classification based on Dense Network (GCDN) for garbage classification to automate and improve the accuracy of this process. Our method utilizes an additional layer of convolutional neural networks (CNNs) to classify garbage into 12 categories such as shoes, green-glass, paper, cardboard, battery, biological, plastic, metal, brown-glass, white-glass and trash. We have executed the different state of the art models of deep learning on a publicly available dataset comprising images of various types of garbage collected from diverse environments. We then employed image augmentation methods followed by transfer learning techniques to fine-tune pre-trained CNN models on this dataset. During the analysis of the results, we have achieved the high classification accuracy of training and validation phase 98.64% and 93.23% respectively. Experimental results demonstrate the effectiveness of our approach in accurately classifying garbage, even in challenging scenarios with diverse backgrounds and lighting conditions. Furthermore, we discuss the potential applications of our system in real-world waste management scenarios, including smart waste bins and recycling facilities, to streamline garbage sorting processes and promote environmental sustainability.

Object Detection in Satellite Images with Canis Hunt Optimized Tetralet Attention enabled Explainable Convolutional Neural Network

2024-06-11T09:48:49+00:00

Object detection has always been a research hotspot in computer vision, specifically detection from satellite images remains a challenging research area. Several conventional researches have been developed but failed to work with high-quality images and satellite images. The object detection in the research is proposed with the Canis Hunt Optimized Tetralet Attention enabled Explainable Convolutional Neural Network (CHunt-TetraExNN). The proposed model aims to provide accurate detection from the satellite image inputs. The Tetralet attention module incorporated with the model is composed of triplet attention followed by positional attention, which provides the accurate estimation of the attentional features that are highly helpful in the process of object detection. Further, the research model is supported by the Canis Hunt Optimization, which is the combination of the adaptability and the hunt characteristics of the Lapins and Latrans that belong to the Canis Family. Thus, the model provided accurate estimation outcomes in the research of object detection from the satellite images, which are estimated with an Accuracy of 96.55%, Precision of 94.59%, Recall of 96.56%, and F1 score of 95.6%.

A Systematic Review of Noninvasive Blood Glucose Estimation Using Near Infrared

2024-06-11T09:51:06+00:00

Diabetes is a chronic and lifelong disease, one of the ten highest causes of death in the world. Diabetes management can only be done by carrying out continuous monitoring. Noninvasive blood glucose measuring devices are needed to overcome the weaknesses of invasive methods, but their accuracy still needs to be improved. This review aims to identify factors that influence the accuracy of estimating blood glucose levels using noninvasive methods based on NIR signals and to observe the development of this technology over the last five years. We performed a systematic review based on articles focusing on noninvasive blood glucose level estimation using near-infrared. This systematic review used the PRISMA 2020 guidelines. Primary studies were retrieved from the literature search engine Scopus database, including journals and proceedings: IEEE, Science Direct, Springer Link, MDPI, Word Scientific, and others. This review provides an overview of using NIR and PPG signals, primary and advanced signal processing, conventional and machine learning approaches, and trends. A total of 62 studies were included. Thirty studies used the conventional approach, and thirty-two studies used machine learning. Thirty-eight studies use primary signal processing, and twenty-four studies use advanced signal processing. Forty studies use NIR signals, and twenty-two studies use PPG signals. India, China, and Indonesia are the top 3 countries in publications on this topic. Using advanced signal processing and feature extraction on photoplethysmography signals and machine learning as an estimation method is quite promising for increasing accuracy. The best machine learning method can be analyzed using meta-analysis.

A DenseU-Net framework for Music Source Separation using Spectrogram Domain Approach

2024-06-11T09:53:36+00:00

Audio source separation has been intensively explored by the research community. Deep learning algorithms aid in creating a neural network model to isolate the different sources present in a music mixture. In this paper, we propose an algorithm to separate the constituent sources present in a music signal mixture using a DenseUNet framework. The conversion of an audio signal into a spectrogram, akin to an image, accentuates the valuable attributes concealed in the time domain signal. Hence, a spectrogram-based model is chosen for the extraction of the target signal. The model incorporates a dense block into the layers of the U-Net structure. The proposed system is trained to extract individual source spectrograms from the mixture spectrogram. An ablation study was performed by replacing the dense block with convolution filters to study the effectiveness of the dense block. The proposed method proves to be more efficient in comparison with other state-of-the-art methods. The experiment results to separate vocals, bass, drums and others show an average SDR of 6.59 dB on the MUSDB database.

Multi-Model Machine Learning Based Path Loss Estimation for Indoor 5G Signal Propagation at 12 GHz

2024-06-11T09:56:19+00:00

In recent years, the advent of 5th Generation (5G) wireless communication technologies has led to a boom in network service usage and access of high-quality multimedia services. In order to maintain acceptable Quality-of-Service (QoS), high-frequency 5G communication at frequencies such as 12 GHz have become commonplace. Consequently, in order to account for signal attenuation, accurate estimation of path loss considering both Line-of-Sight (LOS) and Non-Line-of-Sight (NLOS) propagation, is critical for successful implementation of such wireless communication systems. The present work therefore outlines a multi-model approach to path loss modelling and estimation using standard path loss models such as Close In (CI) and Floating Intercept (FI) models, in conjunction with machine learning (ML) models implementing Random Forest, Decision Tree and Gradient Boosting regression to accurately estimate path loss. The machine learning models employed allow for generation of accurate estimation even in case of significantly varying and noisy datasets. Of the ML models implemented for generation of path loss estimates, the Random Forest Regressor model is illustrated to offer the most accurate and stable results for the given scenario. The results obtained by the multi-model approach are appreciably close to the real-world experimental results, establishing the efficacy of the proposed methods.

Literature Survey on Face Recognition with Hybrid Deep Learning

2024-06-11T10:05:58+00:00

These Face recognition has made remarkable progress with the advent of deep learning techniques. However, accuracy and robustness are still critical for real-world applications. This survey paper explores the synergy between traditional and deep learning methods, providing a comprehensive analysis of hybrid deep learning models for face recognition. We first discuss the foundational techniques in traditional face recognition, such as eigenfaces, local binary patterns (LBP), and histogram of oriented gradients (HOG). These methods laid the groundwork for subsequent developments. We then introduce convolutional neural networks (CNNs), Siamese networks, and FaceNet, which are deep learning models that automate feature extraction from raw facial data.We also discuss the advantages and disadvantages of traditional and deep learning methods, as well as the challenges of hybrid deep learning models. Finally, we present an overview of the state-of-the-art hybrid deep learning models for face recognition. Focus of this survey is the concept of hybridization, where traditional and deep features harmoniously coexist. We provide a detailed examination of key hybrid models, such as DeepID, VGG-Face, and SphereFace, elucidating their architectures, components, and contributions to the field. Additionally, we delve into the integration of face detection and alignment techniques within hybrid models, underlining their significance in achieving accurate and standardized recognition. This paper also presents the latest literature on the Hybrid face recognition models and the techniques used. The paper highlights the advantages of hybrid models, including enhanced robustness, improved accuracy, and computational efficiency, while acknowledging challenges such as data requirements, computational resources, and ethical considerations. It concludes by underscoring the promising future of hybrid deep learning models in elevating the performance and responsible deployment of face recognition systems across various domains, from security and surveillance to human-computer interaction. This survey not only encapsulates the state-of-the-art but also beckons researchers and practitioners to delve deeper into the evolving landscape of face recognition with hybrid deep learning models.

Optimizing SQL Query Execution Time: A Hybrid Approach Using Machine Learning and Deep Learning Technique

2024-06-11T10:08:37+00:00

The escalating volume of global data in recent years has posed significant challenges to data management and analysis, particularly regarding query and processing speeds. In response to these challenges, the present research endeavors to advance large-scale data analytics by accelerating query processing and data retrieval by applying machine learning approaches. The proposed innovative machine learning model aims to improve data retrieval speeds and enhance analytical accuracy. By leveraging the estimated execution time as a guiding metric, the research provides a compass for optimizing query performance. This enables informed decision-making to meet performance requirements and ensures efficient resource utilization within real-time database systems. Notably, the hybrid method introduced in this study demonstrates a reduction in processing time and memory usage, signifying a comprehensive approach to enhancing the efficiency of data management and analysis in the face of burgeoning data volumes.

Identifying Affective Features of Music Tracks to Determine their Popularity using Machine Learning Approach

2024-06-11T10:10:18+00:00

Our world of choices for buying products has often been influenced by our friends, peer groups and now the role of technology in building choices cannot be denied. The songs are an integral part of our life and the choice of songs has very often been influenced by our mood, the song’s digital presence, its lyrics, singer, the band and many more appropriate attributes. The present work is a study of a very popular online streaming app i.e. Spotify that has a strong base of music content and is popular among all the age groups alike. The songs have a set of attributes like accousticness, danceability, energy, instrumentalness and many more that impinge the listener’s mind. The collective set of these features when subjected to the machine learning based techniques bring out the best of the best features out of the songs in the database and that creates a popularity chart. The present work depicts the comparative analysis of the various algorithms for the identification of the popularity. The Random Forest based algorithm shows an accuracy of 80.41%, Logistic Regression with 80.15%, KNN with 77.54% and Decision Tree Classifier shows an accuracy of 68.92%.

Development of WT-ANN Model in thick film SnO2 Gas Sensor for Precise Detection of Volatile Organic Compounds in Exhaled Breath

2024-06-11T10:12:39+00:00

Breath analysis for early-stage detection and monitoring of chronic illnesses, aiming to reduce medical costs and improve patient quality of life. Electronic sensors, functioning as diagnostic tools, can analyze body odors and detect pathological gases. This study focuses on tin oxide (SnO2) thick film gas sensors for detecting VOCs exhaled in breath, including acetone, ethanol, and benzene, which are indicators of diseases like diabetes, lung cancer, and fatty liver disease. A custom gas chamber equipped with a sensor array was constructed, and the sensors' responses to different gas concentrations were recorded. Using artificial neural networks (ANNs), specifically the Wavelet-Transformed ANN (WT-ANN) model, and the study demonstrated the precise detection of VOC concentrations. The WT-ANN employs B-spline wavelet transfer functions for enhanced nonlinearity, allowing for accurate correlation of complex data. Initial results showed that the system could closely estimate acetone concentrations, with minimal error. The findings suggest that the WT-ANN model, combined with semiconductor-based gas sensors, might assist as a non-invasive instrument for diagnosis diseases like diabetes, lung cancer, and fatty liver disease by identifying specific VOC patterns in exhaled breath. The study underscores the potential of ANN-based breath analysis systems in medical diagnostics and highlights the need for continued research to refine this innovative approach.

Optimized Buffer Management Policy for Tailoring DTN Routing Protocols to IoT

2024-06-11T10:15:00+00:00

Delay Tolerant Network (DTN) represents a category of network architectures tailored to challenging network environments. Its primary focus is on addressing network discontinuity, alongside tackling issues like resource constraints and network heterogeneity. Over the past few decades, DTNs have garnered attention as an alternative or complement to existing routing protocols, with a special emphasis on supporting emerging network-based applications that demand enhanced delay tolerance, fault resilience, and flexibility. Among these applications, the Internet of Things (IoT) stands out as a significant domain. This paper provides a brief overview of the commonalities and areas where DTN solutions converge within IoT applications. To enhance delay-tolerant routing in IoT, this work introduces a DTN-based routing protocol known as the Optimised Spray and Wait Protocol (OSnW). This protocol is proposed as a viable alternative for IoT applications with limited buffer resources. Comparative evaluations against three widely used protocols, Epidemic, Spray and Wait, and ProPHET, reveal that the proposed OSnW protocol excels in several key evaluation metrics. The overarching goal of this research is to offer a solution that empowers delay-tolerant routing within the realm of IoT.

AGFC - Augmenting Image Analysis: Gradient Magnitude from a Smoothed Image for Improved Feature Detection in Colorectal Imagery

2024-06-11T10:17:39+00:00

This research presents a comprehensive framework for the processing and classification of multi-modal colorectal images, leveraging an extensive array of data augmentation, neural network models, and advanced techniques. The multi-level classification pipeline commences with a Sequential Convolutional Neural Network (SCNN) and progresses to the subsequent stage, featuring an abnormal tissue detection module incorporating excess object removal and transformers. The architecture further integrates a hybrid Convolutional Neural Network (HCNN), encompassing a Vision Transformer (ViT), a custom cross-modality transformer, a traditional CNN, a Multilayer Perception (MLP), and a combined model. The apex of this approach materializes in a final multi-modal classifier, validating testing images and executing classification tasks. This framework not only showcases a sophisticated and effective strategy for multi-modal colorectal image processing but also exhibits the potential to augment the precision and generalization of Colorectal Cancer (CRC) risk assessments. The incorporation of diverse imaging modalities and advanced neural network architectures positions this method as a robust tool for refining the accuracy of CRC risk predictions in clinical applications.

An Online Exam Proctoring System Using The GMP-DCNN Approach for the Education Sector

2024-06-11T10:19:52+00:00

For the education sector, online examination is an effective tool. However, it has many security issues. Thus, various techniques were developed in prevailing research works. But the performance is still lacking. For solving this issue, a Geometric Mean Pooling-based Deep Convolutional Neural Network (GMP-DCNN)-based Online Exam Proctoring (OEP) system is proposed in this paper. Primarily, video, audio, screen recorder, and app setting screenshots are considered as the input. Next, frame conversion, Kendall Rank Correlated Diamond Search (KRCDS), and Weiner Filter (WF) techniques pre-process the video data. Then, by using the Davies Bouldin Score-based K-Means (DBS-KM) algorithm, the objects are segmented. The face points are identified from the detected objects by using Viola Jones (VJ). Subsequently, the features are extracted from the objects and face points. On the other side, by utilizing WF, the noise is removed from the audio signal. Next, from the noise-removed signal, features are extracted. Next, pre-processing and feature extraction phases are also carried out from the screen recorder. The app setting screenshot was also extracted; from the app setting screenshot, the features were also extracted. By utilizing Schaeffer Weighted Kookaburra Optimization (SWKO), significant features are selected from the extracted features. Next, selected features and all the pre-processed data are inputted to the GMP-DCNN. An alert message is sent to the invigilator if any misbehavior is present. Experimental analysis shows that GMP-DCNN achieves 98.8% accuracy.

Fault Tolerance Enhancement Through Load Balancing Optimization in Cloud Computing

2024-06-11T10:22:26+00:00

In cloud computing, a significant demand for data requests in order to deliver on-demand services at the lowest possible cost. As a result, servers are essential to handle the cloud requests that are dispersed among several geographic zones. Due to less number of servers available in datacenter, some of them are overloaded and some servers are idle or underloaded. This results in requests failing and degrade the system performance. To solve this issue this paper proposed a Particle Swarm Optimization Based Fault Tolerance Load Balancing algorithm (PSOBFTLB). This algorithm is used to provide the flexible and reliability services to each cloud user and maintain the balance of load in each machine by checking the status. To verify our work, a series of experiments over multiple datasets are done by using the CloudSim simulator. According to the simulation results, the PSOBFTLB algorithm works better while using 5% more resources, reduces 15% of the execution time, 12% of the makespan time, 9% of the average response time, and 8% of the average waiting time. Overall, it increases 12% throughput by taking 10% more task is completed as compare with other algorithms such as DLBA and ACO-VMM algorithm.

Enhanced Heart Disease Risk Prediction with Hyperparameter-Tuned Ensemble Models

2024-06-11T10:24:51+00:00

Due to a number of risk factors, heart disease is a serious worldwide health concern that needs quick access to reliable early diagnosis and management techniques. Accurate prediction presents challenges, as seen in the limitations of traditional diagnostic methods. With the growing population, early-stage diagnosis becomes critical. Recent technological advancements have led to research in machine learning applications in healthcare, addressing these challenges. By examining pertinent variables, this work seeks to create an efficient machine learning model for the prediction of heart disease. A number of supervised learning techniques are used, such as XGBoost, K-Nearest Neighbor, Gradient Boosting, Random Forest, Decision Tree, and Logistic Regression. The primary goal is to estimate individuals' heart disease probability based on these factors. In this research, we overcome traditional diagnostic method limitations by utilizing ensemble methods, including the Gradient Boosting algorithm. This approach enhances heart disease prediction accuracy by integrating weak models. These methods open new avenues for heart disease management through detailed data analysis. The results show an impressive overall accuracy score of 99.02%. The developed model provides valuable insights, aiding informed decisions in diagnosis and treatment. Its integration into clinics supports early detection, potentially improving patient outcomes and reducing heart disease-related mortality. Beyond predictions, this study streamlines medical decision-making and revolutionizes heart disease care, enhancing patients' quality of life.

Development of a Web Software for Children and Adolescents Suffering from Autism Spectrum Disorder for Better Learning in Lima, Peru

2024-06-11T10:27:00+00:00

At present it can be seen that many children and adolescents who exercise their classes, whether face-to-face or virtual, have a little difficulty when it comes to learning, in some cases it is because they also suffer from this so-called Autism Spectrum Disorder that in its serious acronym (TEA), for this learning problem the development of Web software is being proposed which can help with the learning of children and adolescents; Through the program to be carried out, it will be possible to provide the necessary help, such as, for example, that they can recognize some of the gestures that their teachers show as well as their classmates. In Peru there are many cases but unlike other countries they chose to be able to implement software that could help them, showing us good results such as the improvement to be able to understand their educational classes, apart from that, being able to recognize the expressions of their peers and also being able to express them themselves; With the proper use of technology aimed at the field of education of children or adolescents who suffer from ASD, we can obtain better results in their learning.

Integrating Augmented Reality and Artificial Intelligence to Enhance the Spatial Ability Skills in the Field of Geography

2024-06-13T09:00:17+00:00

Augmentation Reality (AR) is popular in the field of education, especially as a teaching aid. As evident from the past research, it’s been exhibited that the students found it easier to comprehend complex topics when AR is used. The integration of dynamic spatial data and outdoor AR delivers a solid conceptual basis and technological medium to supply new concepts and innovative access to geospatial visualization depiction. Further, the combination is also dedicated to improving and broadening the participant’s spatial thinking and reasoning capabilities. In addition to AR technology, one of the e-learning platforms, namely chatbot developed using Artificial Intelligence (AI), has become a mainstay in providing long-term support for learning. Thus, this research work nominates an AR-AI framework which was developed as a new way to help middle-school students learn about their surroundings (both inside and out environs). The integration of AR-AI techniques enhances the impact of recognizing dynamic spatial data. This research work sought to meet the academic needs of middle school students in learning their geographic-oriented studies through the proposed AR-AI system.

Mathematical Model of Mastitis Detection Using Milk Data Obtained from Sensors

2024-06-13T09:36:14+00:00

A lot of scientific and research work has been carried out on the detection of mastitis by means of milk data obtained from online sensors of automatic milking systems (AMS) and invasive/non-invasive sensors for animals in livestock farms. In most of these works, mathematical models and algorithms are proposed with different efficiency of mastitis detection as a result of combining certain types of sensor data. However, most farms cannot incorporate enough sensors into their operations due to limited resources and do not conduct laboratory testing activities, which require time, labor and money. This is especially related to animal diseases, which can lead to global problems if timely measures are not taken. In this article, a generalized mathematical model has been developed based on the capabilities of the farm, the effective use of the sensors used in practice, that is, the detection of animal diseases, in particular, mastitis, using sensor data. The originality of the proposed model is that it does not require strict sensor data or indicators related to mastitis. The reason is that, firstly, existing sensor data is processed by linking it to previous historical records, static data, golden rules, and external factors. Secondly, the results of the sensors are summarized by weight coefficients. The result of the model shows the presence of mastitis in the current dairy cow in the [0, 1] interval.

Design and Analysis of Optimal Piped Irrigation Network using CROPWAT, EPANET and Genetic Algorithm

2024-06-14T17:39:17+00:00

Conventionally irrigation water is supplied to the command area through open canals. However, these open channels are subjected to high conveyance losses like evaporation, seepage, and percolation. To reduce these losses Piped Irrigation Network (PIN) is the feasible alternative to Canal Distribution Network (CDN) in the command area. PIN reduces conveyance losses substantially and thereby improves water use efficiency. In this paper, an attempt has been made to design a cost-effective PIN with micro irrigation using CROPWAT, EPANET, and Genetic Algorithm for the Pawale irrigation project, Thane, Maharashtra, India. Environmental Protection Agency Network Evaluation Tool (EPANET) is used to analyze the system, with the required minimum pressure head at demand nodes and velocity as a constraint. Discharge requirement at each demand node is calculated by using CROPWAT. The total cost of the network is the summation of the cost of all pipes in the network. The network with minimum cost is selected with the help of a Genetic Algorithm for the design. It is observed that coupling of CROPWAT model with the EPANET model and optimizing it with GA has the potential to maximize PIN in heterogeneous command areas. This study has indicated that the PIN cost with this approach is approximately 20% less than the conventional design cost.

From Materials to Methods: The Role of Plasmonics in Modern Sensing Applications

2024-06-17T07:25:22+00:00

This systematic review aims to comprehensively explore the evolution of plasmonic sensors from their initial designs to the latest applications. Plasmonics supports the features of sub- wavelength confinement. The most promising aspect of sub-wavelength light transmission is its ability to combine the advantages of optical and nanotechnology. Traditional optical device-based sensors are limited by the diffraction limit, so in this paper, we give a brief overview of the fundamentals of plasmonics. Their various structural configurations, including ring waveguides, silicon-based waveguides, array-based structures, photonic crystal-based structures, and cavity-based structures, have been discussed. The two most popular techniques for plasmon excitation are covered: the first is the wavelength interrogation technique, and the second is the angular interrogation technique. Additionally, a summary of the refractive index sensor's key properties, including sensitivity, resolution, figure of merit, Q-factor, repeatability and accuracy also discussed. It elaborates the properties of many plasmonic materials, including gold and silver. The most crucial element in determining the functioning of the sensor is sensitivity. Therefore, we examine many topologies used in simulate the plasmonics based refractive index sensor, their merits and demerits, their performance parameter, the outcomes of some important works and the author’s critical observation behind the research are analyzed. The unique optical properties of plasmonic nanostructures have enabled the development of sensors that hold immense promise for applications in diverse fields such as medical imaging, environmental monitoring, food safety, and more.

Implementation of 8th Order Delta Sigma Modulation with Reconfigurable Multiple Bandwidth using Truncated Method on FPGA

2024-06-17T07:31:53+00:00

In modern wireless communication technology, an energy and spectrum efficient reconfigurable transmitter design with a high data rate is required; and the conventional systems are wasteful during amplification. For high data rate transmission, LTE-A supports a reconfigurable transmitter design with carrier spacing and continuous carrier aggregation. This study shows how the carrier signal's wide bandwidth is fragmented into several smaller sub-carriers, thus 5G applications need multiband transmission. This work presents an 8th order reconfigurable multi-band delta sigma modulator (RMB - DSM) that enables the noise transfer function zero to be modified to fall at multiple carrier aggregation frequencies. When using several transmission bands, quantization noise between the bands becomes a major issue. Thus, we implement a multi-band additional noise shaping (ANS) function, which greatly reduces on noise over a range of pass-bands by creating notches around each carrier. Systematically designing a 4th order reconfigurable multi-band delta sigma modulator will increase logic size and energy consumption, and the logic's arithmetic operations will need a significant amount of logic in VLSI implementation. The 8th order reconfigurable multi-band delta sigma modulator given in the suggested study is an energy quality scalable truncated technology that would lower the quantity of logic size in arithmetic operations. With these truncated methods, the RMB-DSM architecture's internal and external logic are reduced, and the n x n multiplication yields just n-size output. The proposed approach has been proven by simulation and experiment for aggregating up to four and eight multiband long term evolution (LTE) signals, yielding a total bandwidth of and a sampling frequency of 1 GHz. The Xilinx Zynq 7000 FPGA will be used to implement both existing and proposed designs, and their logic size, latency, and power consumption will be compared.

Performance Graduation Student Predicting Using One-Class Support Vector Machine Algorithm

2024-06-17T08:47:12+00:00

This study explores the prediction of student graduation performance using the One-Class Support Vector Machine (OCSVM) algorithm. The objective is to accurately forecast the time and success rate of students graduating from academic programs. Predicting student performance has become increasingly vital for educational institutions aiming to improve retention rates and support academic planning. The research employs the OCSVM due to its effectiveness in handling imbalanced datasets, which are common in academic performance data. By focusing on a single class, the algorithm can detect anomalies and patterns that signify potential delays or failures in graduation. The dataset comprises various academic and demographic attributes of students from a private university in Indonesia. Data preprocessing techniques such as normalization and transformation were applied to enhance the model's accuracy. The results demonstrate that the OCSVM algorithm can effectively predict student graduation performance with a high degree of accuracy, offering educational institutions a robust tool for early intervention. This approach not only helps in identifying at-risk students but also facilitates the development of targeted support strategies, thereby enhancing overall academic outcomes.

Multi-Model Analysis on Author Attribution Detection on Assamese Text

2024-06-17T08:50:25+00:00

Author attribution detection is a crucial task in the field of forensic linguistics and computational stylometry, aiming to identify the author of a given text based on linguistic features. This study focuses on the application of multi-model analysis for author attribution detection specifically in the context of Assamese text, which is a less explored area compared to other languages. The proposed approach is a first ever attempt for Assamese language, and involves the integration of multiple traditional machine learning models, like Support Vector Machines (SVM), Multinomial Naïve Bayes (MNB) etc. These models are trained on a dataset consisting of a diverse collection of Assamese texts authored by different individual authors. A structured and sizable dataset has been created as part of the current work. Key linguistic features, including word n-grams, character n-grams, and part-of-speech tags, are extracted from the text to represent the writing styles of each author. These features are then used as inputs to the multi-model framework, which combine the predictions of individual models to make a final author attribution decision. Experimental results demonstrate the effectiveness of the proposed multi-model approach in author attribution detection on Assamese text. The study contributes to the Assamese Natural Language Processing, by adding a novel work on authorship detection for these low resources and underrepresented language- Assamese, and highlights the importance of using multiple models for improved performance in computational stylometric analysis.

Maximizing Lifetime of the Network with ML Driven Cluster Head Selection in WSN

2024-06-17T08:52:56+00:00

Cluster head selection is a crucial task in wireless sensor networks (WSNs) for efficient data aggregation and communication. Traditional methods often rely on predefined parameters or heuristics, which may not adapt well to dynamic network conditions. In this study, we propose a novel approach for cluster head selection using machine learning techniques. By leveraging the power of machine learning algorithms, our method aims to dynamically select cluster heads based on various network parameters and environmental factors. We present experimental results demonstrating the effectiveness and efficiency of our approach compared to traditional methods. Our findings suggest that machine learning-based cluster head selection can significantly improve the performance and scalability of WSNs, particularly in dynamic and resource-constrained environments.

A Comparative Analysis of Various Algorithms of Recommender Systems for Serendipity using Novelty Scores

2024-06-17T08:55:21+00:00

The thrust on serendipity is assisting the traditional recommender systems to narrow down on the abundance of recommendations with special weightage and emphasis on waiting-to-be-recommended ‘long tail’ items. Further, it also paves the way for moving from the overlooked ‘accuracy’ aspect of recommender systems to the highly fruitful and rightful aspect of ‘user satisfaction’. As the serendipitous recommender systems inculcate the refreshing ‘novelty’ component, the inherent traditional recommender systems’ issues of ‘long tail problem’, ‘popularity bias’, ‘cold start problem’, ‘over specialization issue’, ‘matthew effect’, etc. are overcome. Hence, in this paper, we investigate and analyze the effectiveness of three different serendipitous recommender system algorithms, TANGENT, KFN and an already published NOVEL SERENDIPITOUS ALGORITHM on a prominent ‘novelty score’ metric. The detailed and rigorous analysis suggest that all the three algorithms are able to surpass the 50 % novelty score benchmark, with the overall novelty scores of 55.57 % for the TANGENT algorithm, 79.39 % for the KFN algorithm and 83.03 % for the NOVEL SERENDIPITOUS ALGORITHM. The results vindicate the overall supremacy and efficacy of NOVEL SERENDIPITOUS ALGORITHM over the other two serendipitous algorithms.

Advanced Image Forensics: Detecting and reconstructing Manipulated Images with Deep Learning.

2024-06-17T08:59:56+00:00

This project presents a comprehensive approach to image forensics, combining deep learning techniques for manipulation detection and image reconstruction. Using Convolutional Neural Networks (CNNs), we accurately classify images as authentic or manipulated, leveraging preprocessing methods like Error Level Analysis (ELA) and wavelet denoising. Additionally, we explore Generative Adversarial Networks (GANs) for image reconstruction, enabling the identification of manipulated regions and assessing alterations' extent. Through experimental evaluation, our approach demonstrates robustness in detecting and analyzing manipulated images, offering a versatile solution for digital forensics and media authentication.

Prediction of Heart Disease using data mining Techniques Based on Hybrid CNN-light GBM Method

2024-06-17T09:03:59+00:00

In today’s situation the heart disease is one of main disease occur among the peoples who are all working in a stressful work Places. A heart disease symptom often requires electrocardiography and blood tests to find accurately; in some complicated task artificial intelligence (AI) provides fast and alternative options.But in this research for finding and predicting heart mortality, morbidity rate we introduce a effective expensive diagnosis process that isdata mining. This data mining performs a unique method for finding the best results in predicting the heart disease like data preprocessing features reduction, data conversion and data scaling are done using the standard dataset in this paper. For the preprocessing the feature scaling method is used for managing the features, variables and the independent range normalizations among the data in the data set. The next step after preprocessing is the feature selection where the features have been selected according to the target variables. For this the recursive feature elimination process has been used for selection where the scanning different feature in the data has been done. After the process of selecting the needed features among the given data’s the process of training and the testing is done for the classification of heart diseases. The work is done using the method of convolution neural networks and with the combination of the Light GBM and predicted using the combination of the above two method as hybrid CNN-Light GBM method for predicting the heart disease patient and health patient. In existing method the around 80% of accuracy has been found among the heart patient data. In this proposed system the existing method has been overcome and found the evaluation metrics performance about 97% of accuracy in finding the heart disease.

The behaviour of patients No-Show in Online Medical Consultation System: A Systematic Literature Review

2024-06-17T09:06:17+00:00

Online healthcare consultation is one of the advents in information and communication technologies (ICTs). Through this, the patient can interact with doctors and attain medical care, including information on community forums, consultations, health records, etc. Over the past few years, patients' popularity for online consultation has increased because of reduced effort. Online medical consultation provides various benefits to patients as compared with face-to-face consultation. Especially, it mitigates several problems hospitals face, such as geographical inconvenience, reduced capacity and long queues. Thus, online medical appointments and consultations highly assist the patient's self-health management. However, the patient no-shows in online appointments can directly influence the services of the healthcare sector. Different related works performed by various authors are summarised in the literature review by referring to several research papers related to the no-show prediction, risk categorisation and online consultation methodologies. A detailed description of different works is provided based on the enhancement of the balancing process to maximise the data handling efficiency and accuracy and to minimise the error rates, model complexities and training time. Also, diverse optimisation strategies are reviewed to enhance the feature selection process with better convergence and prominent feature consideration. By conducting the literature survey, the significance of techniques, the performance obtained, and the drawbacks can be analysed. Through this survey, novel methodologies can be proposed to consider the existing drawbacks to overcome with future directions.

This systematic literature review seeks to improve the thoroughness and transparency of the review process by following a method aligned with the guidelines outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. In accordance with the PRISMA guidelines, this review provides a comprehensive analysis of the existing literature on the behaviour of patients No-Show in Online Medical Consultation Systems. The methodology for the review incorporates a systematic search strategy, stringent selection criteria, meticulous data extraction, and a critical analysis of the findings. The results highlight key trends, themes, and gaps in the literature, while the discussion provides insights, implications, and future research directions. In summary, employing PRISMA in this comprehensive systematic literature review enhances the validity and reliability of the findings, thereby contributing to the progression of knowledge in the domain of patient No-Show behaviour within Online Medical Consultation Systems. With the rising growth of information and communication technologies, an online appointment system is enhanced in several hospitals over the globe. However, an online outpatient appointment system faces various challenges like health, financial, scheduling, and time management problems due to a rising incidence of patient no-shows, referring to patients who do not attend their scheduled appointments. Thus, to assist the hospitals in generating proper decisions and minimise the rate of patient no-show behaviour. The performance metrics are highly utilised to prove the superiority of the machine learning models to predict no-show behaviour in online medical consultation systems.

Friendship Recommendation Algorithms Based on Machine Learning: A Review

2024-06-17T09:08:51+00:00

In today’s digital world, where online social networks are booming, recommender systems play a crucial role in managing the overwhelming amount of information. This paper focuses on friendship recommendation algorithms and their impact on facilitating social connections within the realm of online platforms. It starts by underscoring the importance of recommender systems in alleviating the information overload problem, then delves into an exploration of recommendation algorithms, specifically friending algorithms. The primary focus of the paper revolves around the integration of machine learning techniques into friendship recommendation algorithms, showcasing the potential of artificial intelligence in enhancing social interactions. By combining current research with practical insights, this paper highlights the harmony between machine learning and friendship recommendation algorithms, with the aim of improving personalized and rewarding social experiences in the digital landscape.

Quantum Machine Learning Algorithms for Optimization Problems: Theory, Implementation, and Applications

2024-06-17T09:11:12+00:00

Quantum computing has the potential to transform a number of industries, including machine learning and optimization. This work investigates the relationship between quantum computing and machine learning, with particular attention on the creation, use, and applications of quantum machine learning algorithms for optimization issues. We present a thorough analysis of the theoretical foundation of quantum optimization algorithms, talk about how they are practically implemented on quantum computing platforms, and investigate real-world applications in a several fields. We also highlight upcoming research directions and issues in the realm of quantum machine learning.

Energy and Environmental Performance Enhancement of Radial Four-Stroke Diesel Engines by using Synergistic Staging Configurations

2024-06-17T09:13:54+00:00

The effects of sectioning numbers on a 1500 RPM radial six-cylinder diesel engine with a 150 mm bore, 180 mm piston stroke, and a 15:1 compression ratio is examined in this study. The results of tests conducted under typical conditions (288 K, 1 bar) were evaluated for sectioning numbers ranging from 2 to 10. According to the specifications of the intake system, the gas velocity went from 15 m/s to 80 m/s, but the average intake manifold pressure stayed the same at 1.9 bar. The air-fuel ratio, cylinder pressure-temperature angles, and swirl ratios were all kept constant. As the number of sections increased linearly, the maximum gas force acting on the piston also increased, suggesting an improvement in performance. Improvements in combustion were indicated by an increase in exhaust gas temperature from 746 K to 775 K and a drop in average exhaust manifold gas pressure due to lower sectioning numbers, according to the exhaust parameters. According to ecological research, the best environmental performance is achieved with a sectioning number of 10, and emissions decrease as the number of sections increases. Complying with air pollution rules and demonstrating the advantages of sectioning modifications for increased engine performance and sustainability, this design decreased emissions.

QoS Achieving Sector Based Congestion Control Protocol in Wireless Sensor Networks

2024-06-17T09:15:49+00:00

The current innovation pertains to an info acquiring mechanism for wireless sensor networks. The goal is to enhance Level of Performance in wireless sensor networks with applying bandwidth management sectoring techniques that decrease sensor node usage of energy. The nodes are distributed periodically throughout the system. There are several sectors constructed based on a comparable amount of sector heads. There is a single sink node that takes information from the sector leaders. Common nodes are distributed at randomly to transfer information to the corresponding level 1 node across a sector. Thus, channels for communication are managed, and network activity is controlled. The sector-based congestion reduction method fulfils different quality-of-service criteria for wireless sensor networks. Qualities of Service indicators include time (delay), energy utilization, delivery ratio, loss ratio, and throughput. This study describes the optimal methodology for accurate information distribution in wireless sensor networks. This is intended to achieve optimum QoS for wireless sensor networks. We have achieved QoS of WSN’s like PDR, PLR, Throughput, Energy efficiency, Delay etc. This paper will be useful for new a researcher who wants to work on congestion control parameters of real time applications.

Hybrid Neural Network with Weighted Modified Cuckoo Search Optimization for Software Defect Prediction: A Soft Computing Approach

2024-06-17T09:17:46+00:00

Defects in software significantly impact quality, reliability, and maintenance. Early detection and prediction using data mining and classification techniques offers an effective means of identifying potential defects before they manifest in production environments, but accurate prediction requires handling complex datasets. This paper proposes a soft computing model called Hybrid Neural Network with Weighted Modified Cuckoo Search Optimization (WMCSO) to detect the defect in the software. The proposed model first performs the clustering process with the Modified Fuzzy C–means algorithm (MFCM) to retrieve the important new attributes from the dataset. The software defect prediction and classification are performed using the HNN, and the WMCSO model is used to fine-tune the weights of the HNN. The HNN-WMCSO method is evaluated based on the evaluation of prediction rate and execution time. The experimental analysis stated that the proposed model exhibits improved performance relative to the current method in regard to an efficient prediction rate.

Time Allocation to Management Activities of Retail Msmes Through an Intelligent System

2024-06-17T09:21:22+00:00

Effectively handling a manager’s time is critical for organizational success. This paper describes a methodology that uses cutting-edge decision support to offer flexible recommendations grounded in the unique context of Ciudad Juárez, México. The methodology incorporates manager preferences, determining the value of activities through the theory of value functions and performs an optimization based on differential evolution. This methodology ensures practical applicability, handling multiple activities, group decisions, and uncertainties prevalent in the dynamic business landscape. Validating the approach, a real case study involving Ciudad Juárez managers provides contextual evidence. The proposed system’s performance is evaluated based on the satisfaction levels of participating managers. This research contributes to decision and social sciences by introducing a system designed for handling managerial times in a culturally specific environment. The study emphasizes the adaptability of the system to diverse managerial activities, making it a valuable tool for decision-makers.

An Innovative Method: Identifying Pests Through Artificial Neural Networks and Image Processing

2024-06-17T09:23:41+00:00

Effective and precise pest detection technologies are of the utmost importance on a global scale in order to reduce the negative effects of pests on crop yields. “This research presents a new method for detecting pests using ANNs and image processing tools. By combining machine learning with image analysis”, our suggested approach offers a powerful tool for pest identification in a variety of agricultural contexts. Artificial neural networks (ANNs) make it possible to train specialised models that can visually distinguish between different kinds of pests in digital photos. Research using large real-world datasets has shown that preprocessing techniques improve model performance and feature extraction, making them more efficient and accurate than traditional pest detection methods. This research contributes to the field of precision agriculture by providing a trustworthy and automated method for early pest detection, which allows for prompt action and reduces crop loss. Our method takes use of ANNs—which can learn complex patterns from picture data—by combining the most recent developments in deep learning with image processing. Morphological operations and histogram equalisation are two preprocessing methods that help minimise noise and improve the discriminative power of retrieved features. Our technique has been rigorously tested across multiple datasets with different pest species and habitats. It has proven to be quite accurate and scalable in agricultural settings. “The automation and efficiency benefits of our technology are further highlighted when compared with traditional pest identification methods, such as chemical-based procedures and human inspection”. This research highlights the significant impact that AI and image processing may have on pest control tactics. It paves the door for agricultural systems that are more robust and sustainable, and can better handle new threats as they arise.

Investigation of Suitable Anti Spoofing Algorithm for GNSS Receivers

2024-06-17T09:26:15+00:00

In an era driven by transformative technologies like 5G and the Internet of things (IoT), the Global position system vulnerability)-based navigation systems to spoofing attacks has become a paramount concern. Various techniques have been proposed to enable detections .This paper explains about anti –spoofing algorithms such as adaptive Kalman filter for dynamic positions and in GPS navigation, situations may arise where GPS receivers lack a clear line of sight to enough satellites ,such as when they are inside buildings .tunnels ,or aircraft with limited sky view .To tackle this challenge ,there is a techniques for initializing GPS receivers’ to rapidly and effectively track signals once they are regain access to a clear sky vie .Furthermore ,in the context of unmanned aerial vehicles (UAVs) vulnerable to deliberate interference like spoofing attacks ,we present a comprehensive solution for GPS Spoofing detection and mitigation. This approach involves distributed radar ground stations equipped with local trackers .connected to fusion node

Implementation of Inertial Measurement Unit (IMU) Sensor for Monitoring System Motion Monitoring System for Pregnant Women (SIMBUMIL)

2024-06-17T09:29:32+00:00

Pregnancy is an important period of life until the birth of the baby in the womb. One of the biggest concerns in the first trimester of pregnancy is the occurrence of miscarriage caused by excessive activity. This research examines the implementation of an Inertial Measurement Unit (IMU) Sensor using IMU MPU-6050 for pregnant women's movement monitoring system (SIMBUMIL). The movement data of pregnant women is monitored in the form of movements from daily activities. The SIMBUMIL prototype device is attached to the arms and legs to obtain real-time movement data. Based on the test results, it shows that the device can detect any angle changes experienced by pregnant women during activities. The average amount of data sent and stored in the database is 7 movement data per second per device. The SIMBUMIL application also successfully displays the movement data of pregnant women's activities visually in the form of graphs.

A Review of Unmanned Aerial Vehicles/Urban Air Mobility Potential Cyberattacks and Authentication Models: A Way Forward

2024-06-17T09:32:15+00:00

An Unmanned Aerial Vehicle (UAV) is an aircraft that operates without an onboard human pilot. It can be remotely controlled through a ground control station (GCS), remote control, or onboard computer programs. The elements onboard use a network of sensors to communicate with GCS via a wireless link and thus make the system susceptible to various cyber-attacks. These have magnified concerns, especially in recent years, due to the increased adoption of drones across multiple sectors such as governments, industries, businesses, and transport. Given the paramount need to ensure availability, integrity, and confidentiality, securing these systems is crucial. The attacks may present in forms such as jamming, denial of service, signal attack, eavesdropping, hijacking, man-in-the-middle, intrusion, and malicious application, among others, and these attacks could be mitigated using an effective authentication model. This research reviews several such models, majorly cryptographic, lightweight, and blockchain-based, proposed by different scholars. Considering the importance of blockchain, this research grouped these authentication techniques into two: blockchain and non-blockchain-based. The study shows that all the reviewed authentication techniques have certain limitations, indicating the need for enhancement. Finally, this review identifies the need to consider UAV's peculiarities, operating environment, communication channels, energy consumption (battery life), and blockchain technology to formulate an optimal authentication model.

Detecting Diabetic Retinopathy using Deep Learning

2024-06-17T10:44:32+00:00

Diabetic retinopathy (DR) is a leading cause of vision loss in working-age adults worldwide, and timely diagnosis is crucial for preventing blindness. Current methods for diagnosing DR rely on manual grading of retinal images, which is time-consuming and prone to inter-observer variability. The development of DR is a complex process involving multiple cellular and molecular pathways, including inflammation, oxidative stress, and vascular dysfunction. Despite its significant impact on public health, there is currently no effective treatment for DR that can halt or reverse its progression. Recent advances in deep learning and image processing have opened up new possibilities for automating the detection of DR. The aim of this study is to develop a system that can accurately classify individuals suffering from diabetic retinopathy. Filtering algorithm is used to clean and preprocess the images collected by users, thereby ensuring the accuracy of the results and reducing the impact of noise on the diagnostic process. An efficient custom three layer CNN model with hyper-parameter tuning is used on kaggle ‘ilovescience’ dataset which gives promising accuracy of 94.45%.

Navigating the Scalability Maze in Blockchain Technology – An Analysis of Scalability, Transaction Issues and Solutions

2024-06-17T10:46:38+00:00

Recently, blockchain has garnered immense attention from both public and private sectors as the most sought-after technology. Despite its potential, scalability remains a critical challenge hindering its full realization. Blockchain provides a secure and transparent network through its features such as trust, data security, decentralization, immutability, and transparency. The potential of blockchain technology to revolutionise numerous industries has attracted a lot of attention in recent years. Scalability, however, is among the major issues preventing its mainstream use. Blockchain networks encounter capacity and speed constraints as they expand in size and complexity, which results in higher transaction fees and delays. The scalability problems with blockchain technology are examined in this paper, along with a thorough discussion of the numerous scaling solutions put out in the literature. The solutions for scalability and transaction speed can be categorized into two main groups - on-chain and off-chain. On-chain solutions include Segwit, block size increase, Sharding, Directed Acyclic Graph and Consensus mechanisms, while off-chain options encompass Interoperability technique includes payment channels, cross-chains, and side-chains and Finally, in this paper, we have covered well-known scalable consensus mechanisms and the future directions of block chain in terms of scalability and transaction throughput.

Deep Learning-Based Group Activity Recognition Using Multiperson Relational Graph

2024-06-17T10:52:28+00:00

The recognition of group activities (GAR) is of significant importance in the field of computer vision as it facilitates the investigation and understanding of patterns of human behavior. Existing methodologies mostly concentrate on interactions at the interpersonal level within a group. However, sociological research has emphasized the significance of individual characteristics, interactions at the multi-person level, and the overall structure of the group in recognizing group activities. Hence, in this research, to represent the relationships between people’s locations and appearances, adaptable and effective multi-person relational graphs (MRG) have been developed for the aim of GAR. Graph Convolution Network (GCN) with sparse temporal sampling is applied to efficiently infer multi-person relational graphs. The proposed network distinguishes group activity from individual interaction via relational reasoning. The use of a GCN for identifying group activities comes after the implementation of a deformable CNN to collect features and categorize individual actions. For multi-level interaction reasoning and group structure modeling, visualization samples and experimental results show that this approach works better than the best methods currently available. These findings highlight the necessity of taking into account multi-person relational graphs (MRG) representations for recognizing group activities.

Predictive Modeling of Dropout in MOOCs Using Machine Learning Techniques

2024-06-17T10:54:34+00:00

The advent of Massive Open Online Courses (MOOCs) has revolutionized education, offering unprecedented access to high-quality learning materials globally. However, high dropout rates pose significant challenges to realizing the full potential of MOOCs. This study explores machine learning techniques for predicting student dropout in MOOCs, utilizing the Open University Learning Analytics Dataset (OULAD). Seven algorithms, including decision tree, random forest, Gaussian naïve Bayes, AdaBoost Classifier, Extra Tree Classifier, XGBoost Classifier, and Multilayer Perceptron (MLP), are employed to predict student outcomes and dropout probabilities. The XGBoost classifier emerges as the top performer, achieving 87% accuracy in pass/fail prediction and 86% accuracy in dropout prediction. Additionally, the study proposes personalized interventions based on individual dropout probabilities to enhance student retention. The findings underscore the potential of machine learning in addressing dropout challenges in MOOCs and offer insights for instructors and educational institutions to proactively support at-risk students.

Graph-Based Approach for Solubility Prediction of Drugs using SMILES Data

2024-06-17T10:56:42+00:00

Graph Neural Networks (GNN) utilization in the case of molecular property prediction is considered a significant advancement in computational chemistry and drug discovery. Initial approaches to molecular property prediction especially solubility prediction depend on empirical rules or physicochemical descriptors, which lack generalization and predictive accuracy. The proposed model Graph Convolutional Network (GCN) which is a variant of GNN learns representations of molecular graphs, enabling accurate prediction of molecular properties directly from raw molecular structures. The molecular graphs are created from the Simplified Molecular Input Line Entry System (SMILES) data which are molecular sequences of drug target compounds. In the proposed work, GCN uses graph pooling, which effectively reduces the node dimensionality. This work shows how the whole graph can be considered as input and how different pooling techniques can be used to handle large and complex graph data and also the effectiveness of GCN for solubility prediction. The proposed GCN model is hyperparameter tuned by using Grid Hyperparameter optimization on ESoL dataset which is a regressive type dataset achieving a low RMSE value of 0.43 outperforming machine learning and many deep learning models.

An Exhaustive Investigation into Energy Preservation and Cutting-Edge Technology Utilized in Cloud Data Centers

2024-06-17T10:58:55+00:00

The rapid growth of the digital economy is a result of the data center's rising worldwide energy usage. Data centers are thought to be the hub with the highest energy use. The public is paying close attention to information centers in an effort to cut down on energy emissions and meet energy consumption targets. Enhancing the Cloud data centers energy efficiency is a significant area of a fascination with the scientific community. Researchers are putting an abundance of effort into implementing numerous measures and a feasible energy efficiency implementation approach to be able to meet these goals. In this study, we categorize the current energy efficiency measures, provide an overview of the approaches, and utilized to assess data centers' energy efficiency. In this paper, we look at the current situation and difficulties in assessing the energy efficiency of data centers and provide a survey for enhancing energy efficiency assessment tools to help cloud operators. Through the use of more sophisticated metrics to access advanced data center energy efficiency, this work provides academics and decision-makers with ideas for building appropriate ways for evaluating energy efficiency. It also encourages them to connect theory and practice in energy efficiency evaluation. It is the most substantial and critical step toward achieving sustainable development goals and cutting edge green technologies.

Brain Tumor Segmentation and Classification in Mr Images Using Residual U-Net Semantic Segmentation Model

2024-06-17T11:01:05+00:00

Brain tumor identification makes use of machine learning and computer vision methods in order to automatically identify and categorize brain cancers in medical images such as MRI scans. A model for the segmentation and identification of brain tumors based on a residual U-Net is proposed in this work. The Residual U-Net is an altered variant of the U-Net architecture that is used in semantic segmentation tasks. It incorporates the concept of residual connections, which allows for the model to learn more complex representations of the input data and can result in more accurate segmentation. Residual connections allow for deeper networks to be trained while mitigating the vanishing gradient problem. This resulted in accurate segmentation of brain tumors, particularly in cases where the tumors are small or located in complex regions of the brain. The proposed model obtained a higher tversky of 0.89 and higher PSNR of 30.

Inheritance and Updating Strategies of B&B Design Concepts in the Context of Artificial Intelligence-A Case Study of B&B Design in Southern Anhui, China

2024-06-17T11:03:19+00:00

These instructions address the issues of serious homogenization, regional weakening, lack of systematic renovation, and the inheritance and protection of traditional culture in current rural guesthouse designs in southern Anhui Province. It explores optimization strategies for these designs from multidisciplinary perspectives such as architecture, geography, and design. The study integrates the viewpoints of design experts, consumers, and operators, proposing an innovative design strategy. Using a mixed qualitative and quantitative approach and drawing on existing literature as theoretical basis, it emphasizes the regional culture of southern Anhui. Through questionnaires and interviews with designers, consumers, and operators, it investigates innovative design strategies for guesthouses under the background of artificial intelligence. These strategies not only replicate previous consumer response models but also introduce innovative design patterns and insights, providing a comprehensive framework for the design strategy of rural guesthouses in southern Anhui. The research establishes a theoretical basis and practical guidance for guesthouse design based on regional culture, emphasizing the integrity of design and consumer interaction. It contributes to the output of regional culture, the development of the guesthouse industry, and the enhancement of market competitiveness. The study proposes future strategic research directions, further exploring the integration of artificial intelligence and the distinctive regional cultural features of southern Anhui to meet the diverse consumption demands and aesthetic preferences in guesthouse design driven by the booming tourism industry.

Detecting Deepfakes: Exploring Machine Learning Models for Audio, Video, and Image Analysis

2024-06-18T07:46:12+00:00

The rapid evolution of deepfake technology has created substantial hurdles for the detection of altered media. This study investigates the field of deepfake detection with an emphasis on the use of machine learning techniques in the fields of image, video, and audio analysis. The effectiveness of several machine learning models—Random Forests, Gradient Boosting Machines, Support Vector Machines, Neural Networks, and Convolutional Neural Networks, among others—in identifying deepfakes is compared and contrasted. The analysis outlines the benefits and drawbacks of each model and offers performance insights derived from real-world case studies and research findings. The paper also addresses recent developments in deepfake detection techniques, including ensemble learning approaches and ResNet topologies, which present interesting directions for further research and development in the fight against the spread of manipulated media.

An Innovative Approach to Enhance the Safety of Elevator Using Steel Cable Damage Detection Model Based on YOLO

2024-06-19T02:59:43+00:00

To address the issues such as limited detection device resources and prolonged detection times in surface damage detection of steel cables installed commercial, public, and industrial buildings, advanced deep learning techniques, and Convolutional Neural Networks (CNN) have been investigated in this study and a new network model has been designed. This work proposes a steel cable defect detection network model based on YOLO, incorporating GhostNet into the backbone network, and introducing a novel feature extraction module (ShuffleNC3) based on ShuffleNet and attention mechanisms. Pruning improvements are then applied to the Head part. Experimental results indicate that the improved network achieves approximately1.149% increase in average precision compared to the baseline YOLOv5s. This modification achieves a simultaneous reduction of network computational costs and maintains high recognition accuracy, meeting better requirements for surface damage detection in steel cables. The parameters and computational costs are reduced by approximately 43 % and 31.4%, respectively, while the model size also decreases by 42%.

Digital & Social Media Impressions in Khadi Industries: An Effective Tool for Global Reach, Sales and Awareness of Khadi

2024-06-19T07:36:07+00:00

Purpose – The paper aims to indicate the importance of accommodating the incumbent legacy khadi institutions by using the digital social media platforms for E-commerce solutions. The paper gives an in-depth study of the various leading fashion industry brands which have incorporated the social media platform for their growth success in e-commerce marketing and wider reach. The research paper studies the significant features of the digital marketing and aims to assist Khadi industries and Khadi Institutions to integrate digital advertising and marketing in their business models to recognize the usefulness of social media marketing platforms. This paper focuses on the inherent usage of digitalisation in the khadi industries.

Design/methodology/approach – The paper draws references from the Karnataka zonal khadi institutions which have incorporated E-commerce. The paper studies the Digital Mark Impressions of the KVIC in social platforms. The paper analyses a deep study about the social media platform reaches and success of some popular fashion brands. And concludes that the digital mark in the social media platforms progressively assist for the higher growth in sales and popularity of the Khadi product.

Findings – In the Research Exploration process of digitalization, the study concludes that there is a progressive interrelation exists between the Digital Mark of KVIC in social platforms and the Performance Growth of the Zonal Khadi Institutions with relation to sales and popularity. The comparative study concludes that the performance of the Karnataka Zonal Khadi Institutions increased and the Swadeshi Khadi fabric achieved higher sales performance and popularity due to the predominance mark of MSME, KVIC on the Digital Social Media Platform. The paper gives an in-depth study of the various leading fashion industry brands which have incorporated the social media platform for their growth success in e-commerce marketing and wider reach.

Originality/Value – The paper suggests that Digital Social Platforms and Digital Operations in Khadi Institutions acts as successful factors on the higher performance and higher awareness about the Heritage Fabric. The study adds new evidence on the policy approaches for expanding the access to the digital websites and social platform for all the Khadi Institutions. The study suggests for the government policies on the compulsory production and purchases of the Khadi Fabric across PAN India by all retail, wholesale, online and offline store outlets to showcase Khadi for sales in their catalogues.

Congestion Avoiding Approach Using Optimization Algorithm for IoT Web Services

2024-06-21T07:01:16+00:00

The Internet of Things, or IoT, is a technology that uses real-time data transfer via sensors to monitor remote sites and other systems. In this study, the "Thingspeak" web service, which is based on the Internet of Things, is examined as an open access API service that serves as a host for a range of sensors to monitor data received from sensors at the cloud level and transfer the data to the MATLAB platform at the designated channel ID with an API key. It is suggested to use an alternative optimization strategy for handling congestion and load balancing with Thingspeak-based IoT web services. This would improve network performance by distributing traffic load along the best paths possible to prevent traffic congestion.

A Note on the String Metric for Word Similarity

2024-06-21T07:03:50+00:00

This paper presents a string metric for measuring the similarity between words. The distance function satisfies the axioms of non-negativity, reflexivity, symmetry, and triangle inequality. A comparative study of the string metric is carried out with Hamming and Levenshtein distances for word matching task.

GridDR: Enhancing Grid Reliability using Demand Response Program

2024-06-21T07:07:59+00:00

In developing nations, power disruptions are a major worry, and grid stability is essential. Utilities must encourage energy consumption reductions by consumers during prime hours in to achieve and maintain grid stability and avoid brownouts or blackouts. Finding suitable candidates for Demand Response (DR) events is essential. . In order to strategically select candidates for DR events based on the utility's goals, this work suggested "GridDR," which gives users the ability to monitor their energy usage trends, customize their choices for participation, and receive tailored advice or incentives for taking part in demand response. In addition, the platform offers distributors thorough visualizations of customer energy usage data, allowing for the early identification of high-usage customers for demand response involvement. The study makes use of a dataset that includes hourly energy usage data gathered over a one-year period from 39 apartments to assess consumption trends and find possible participants in demand response The study starts with a thorough project overview, emphasizing the importance of demand response programs in resolving grid stability and reliability issues. With the intention of offering insights into temporal fluctuations and consumption trends, graphical analytic techniques are used to show daily, weekly, and monthly energy use patterns based on the dataset. Subsequently, two clustering algorithms, namely K-means and hierarchical clustering are used in this research work. GridDR has the potential to completely change how distributors and customers communicate and work together to optimize energy use and improve grid reliability by bridging the gap between data analytics, user interface design, and demand response program execution. In the end, the study emphasizes how critical, is to employ innovative approaches to leverage data-driven insights for the purpose of managing the changing issues of grid reliability and energy management in the residential sector.

A Bibliometric Analysis of Soil Nutrient Testing Using Digital Image Processing with Vosviewer

2024-06-21T07:16:59+00:00

As we know our environment is deteriorating rapidly. We need technological interventions to speed up the work done in order to save our environment. Our scientists, researchers, agriculturists and government invest a great deal of money in order to test soil samples. This area of research is based on use of technology in order to help with soil testing. This study is done to examine current and future trends in research related to “soil nutrient analysis using image processing”. We have used Scopus database to get articles related to area of interest and VOSviewer (1.6.19) version is used to extract details from articles using bibliometric analysis. Parameters such as keywords, author names, publication journals, organization, countries and citations have been considered. Statistics show that soil nutrient analysis using machine learning is a trending topic in research field. A total of 1400 publications were used from Scopus database. Findings suggest that maximum research has been done in this field by researchers of China. Major work has been done on identifying soil calcite, soil organic carbon etc. using image processing and minor work has been done in order to identify other micro-nutrients, macro-nutrients in soil such as NPK and pH of soil.

The Analysis of Factors Affecting the Successful Implementation of a Forest Management Information System in Forestry Company

2024-06-21T07:19:37+00:00

The present study offers a comprehensive review of the determinants of the success of an Indonesian forestry company's implementation of the Forest Management Information System (FMIS). By integrating variables from multiple prior works into an updated version of the DeLone and McLean model, this study investigates six key variables: System Quality, Net Benefits, User Satisfaction, Information Quality, Information Use, and User Quality. Enhancing the efficiency of FMIS deployment and minimizing potential failures in subsequent implementations are the principal objectives. In order to acquire data, a questionnaire that had been carefully crafted was utilized, guaranteeing that responses would be representative and dependable. The refined DeLone and McLean model played a pivotal role in identifying critical determinants that underpin the achievement of information systems objectives. The results unequivocally demonstrate the efficacy of the FMIS deployment at the forestry company in Indonesia, with User Satisfaction emerging as the most critical determinant with a mean score of 0.859. Information Quality, User Quality, Information Use, and System Quality follow in close succession. The study substantiates the substantial and favorable influence of these variables on user contentment and the overall net advantages, providing valuable perspectives on the effective execution of information systems, specifically within the framework of FMIS.

Performance of WBAN under Different Matrices with Mobile Sinks

2024-06-21T07:21:44+00:00

From last few years, several models for postural mobility signifies in successful deployment in form of outputs. Because the QoS based performance of WBAN are extensively controlled of different topologies. The different networks nodes are connected with other nodes with their movement pattern in different network according to this research work. In this research we observe the performance of WBAN under four performance matrices as throughput, packet delivery fraction or PDR, consumption of energy and EED. FANET is a domain and use mobile sink as a successful outcome. By using Network Simulator NS2.35 for different mobility model were tested. And finally, we obtained as mobile sink improved the performance of WBAN under different performance matrices.

Analyzing the Bit Error Rate of NOMA over Rayleigh Fading Model.

2024-06-21T07:25:56+00:00

Our work investigates the performance characteristics of Non-Orthogonal Multiple Access (NOMA) in the context of wireless communication systems that are affected by Rayleigh fading channels. Here the characteristic performance is nothing but calculating the Bit-Error Rate (BER), capacity, and outage probability of the wireless communication system in the integration of Non-Orthogonal Multiple Access (NOMA). Non-orthogonal multiple Access (NOMA) is one of the multiple access schemes that aims at enhancing the efficiency of the spectrum. It also ensures that multiple users can accommodate the same frequency band. In this research, we will be developing a network of two users: a near user and a far user. By utilizing the MATLAB simulations, we can access the Bit-Error Rate (BER), capacity, and outage probability. Here we will have a base station from which a down-link transmission is connected to the users. To uphold fairness across the users, power allocation factors are duly managed. The main reason behind the addition of Rayleigh’s fading effect is, that it enables the realistic behavior of the channel. In Rayleigh’s fading, each transmitted bit will somehow encounter varying degrees of attenuation and phase shifts due to the multipath phenomenon. Simulation outcomes shed light on the ramifications of altering transmit power levels on key system performance metrics, thus providing valuable insights into the intricate interplay between Bit-Error Rate (BER), capacity, and outage probability. The above findings will play a vital role in efficient power allocation strategies and Successive Interference Cancellation (SIC) techniques in increasing the efficiency of NOMA. Overall, this paper study contributes to a deeper comprehension of NOMA’s operational dynamics within real-world wireless communication environments and also offers some significant implications for future system design.

Evaluating the Software Engineering Curriculum at JUST: A Comparative Analysis with IEEE Knowledge Areas

2024-06-21T07:28:42+00:00

Contribution: This paper examines the compliance of the Software Engineering (SE) program at Jordan University of Science and Technology (JUST) with the fifteen Software Engineering Knowledge Areas (SEKAs) established in the IEEE-CS SWEBOK. This research is the first to determine the scope of these knowledge fields in a SE undergraduate program curriculum.

Background: Although the Institute of Engineering and Technology (IET) offers the SE undergraduate program at JUST, aligning it with the IEEE view of SE is essential.

Research Questions: This research aims to answer some questions, such as how much the SEKAs’ topics are included in JUST's SE undergraduate program. How to eliminate the identified coverage gap.

Methodology: Observe the coverage of the SEKAs' topics in the SE undergraduate program curriculum courses at JUST. Then, accordingly CLASSIFY theSEKAs into (a) Full-Compliance (e.g., when concerned Knowledge Area is fully covered across one or more of the SE undergraduate program curriculum' courses); (b) High-Compliance (e.g., when concerned Knowledge Area is highly covered); (c) Partial-Compliance (e.g., when concerned Knowledge Area is partially covered); or (d) Poor-Compliance (e.g., when concerned Knowledge Area is poorly covered).

Findings: This research concluded that the SE undergraduate program complies with the Software Requirements (SW-RQTs), Software Testing (SW-TS), SE Management (SW-MG), and Software Quality Knowledge Areas (SQKAs). The Software Design (SW-DS), SE Models and Methods, Computing Foundations (CFs), and Mathematical Foundations Knowledge Areas (MFKAs). c) Ensure that the software construction (SW-CN), maintenance (SW-MN), configuration management (SW-CM), and SE process knowledge areas are consistent and compliant. d) Unsatisfactory in the SE Professional Practice, SE Economics, and Engineering Foundation Knowledge Areas (EFKAs).

Elastic Data Analytics in Healthcare: Enhancing Patient Outcomes

2024-06-21T07:31:32+00:00

Effective management and study of Electronic Health Record (EHR) statistics are essential in transforming healthcare delivery and decision-making. This paper explores the amalgamation of Elastic Data Analytics (EDA) into EHR systems as an approach to address the tasks of managing vast volumes of heterogeneous healthcare data. EDA enables healthcare groups to dynamically scale data resources and adapt analytics workflows to changing requirements, facilitating real-time access to information and data-driven decision-making. The application of EDA in specific healthcare domains, such as cardiovascular disease (CVD) monitoring, highlighting its prospective to reform disease prediction, population health management, and personalized medicine approaches. However, successful implementation requires addressing tasks such as data interoperability, privacy and security concerns, and scalability of healthcare infrastructure. Cooperative efforts between healthcare services providers, data scientists, and policymakers are essential to harness the full potential of EDA and drive positive outcomes in healthcare delivery. This paper underscores the transformative impact of EDA in healthcare and provides visions into its future suggestions for improving patient care and advancing healthcare innovation.

Powered Hazard Distribution through Order Statistics and Its Applications

2024-06-21T07:34:05+00:00

This article addresses the distribution of order statistics of the power hazard distribution with graphical and quantitative measures along with cumulative residual entropy. We study the single and double moments and establish the recurrence relation between them. Finally, real data is analyzed to show the usefulness of our results.

Enhancing Cardiovascular Disease Diagnosis: A Hybrid Model of Whale Optimization Algorithm with Multilayer Deep Perceptive Classifier

2024-06-21T07:35:58+00:00

An efficient Hybridization of Whale Optimized MultiLayer Deep Perceptive Classifier (HWO-MLDPC) is proposed to improve the diagnosis accuracy of cardiovascular disease. The proposed technique includes three main stages: preprocessing, feature selection, and classification. First, the data is preprocessed using the Theil-Sen Regressive Discretized Binning method, which smoothes the raw data into a structured format based on median estimation. After preprocessing, the feature selection process uses stochastic bivariate correlation to identify relevant features based on maximal mutual information. Next, classification with the selected relevant features is performed using the Hybridization of Whale Optimized MultiLayer Deep Perceptive Classifier. The proposed MultiLayer Deep Perceptive Classifier comprises several layers. First, the number of selected features is given to the input layer. Then, the input is transferred to the hidden layer, where feature analysis is performed using the Generalized Tversky index similarity. The sigmoid activation function provides the final disease classification results. At the same time, whale optimization updates the weights of inputs with lesser error to achieve accurate classification results with minimum error at the output layer. Based on the classification results, cardiovascular disease can be diagnosed correctly. Experimental evaluation is carried out using different quantitative metrics such as accuracy, precision, recall, F-measure, and time complexity. The analyzed results demonstrate the superior performance of the proposed technique.

IDS Range of Attack in WSN and Prevention Using PSO Fitness Measures

2024-06-21T07:44:30+00:00

The concept of Euclidean space is used in sensor ad-hoc network routing techniques. This system's methodologies will derive the position of sensor nodes from previously used distances that measure static and dynamic sensor node placement. When the position of a node is known ahead of time, the number of security vulnerabilities increases. To put it another way, our strategy is to use PSO to detect security breaches and create an effective Intrusion Detection System. The simulation demonstrates the effectiveness of the DREAM protocol in thwarting an Intrusion Detection System attack aimed at duplicating sensor IDs within boundary areas. It's assumed that the sensor boundary maintains records of sensor nodes and their neighboring nodes, along with the distances between them, as dictated by the routing protocols. We assume that the sensor border maintains information about the sensor node and its neighboring nodes, along with the distances between them as dictated by the routing protocol. By employing PSO, we achieve a fitness value that tends to position the sensor node's neighbors within the border, thereby enhancing the secure zone for ad-hoc transmission.

Sentiment Analysis for Prediction of Brand Value Using Albert Model

2024-06-21T07:47:05+00:00

In today's dynamic and global business environment, organizations face the challenge of meeting customer expectations while effectively managing their supply chain. Understanding customer demands and accurately getting product sales is critical to achieve this. The effectiveness of the existing “BERT” in sentiment analysis is well established and its resource-intensive models might face challenges in deployment, especially in scenarios with constraints on computational resources. This study explores the use of sentiment analysis and the ALBERT model to predict brand value based on customer reviews. Both BERT and ALBERT models are more powerful, but ALBERT offers a more efficient alternative without compromising performance, making it particularly appealing for tasks where computational efficiency is a priority. The proposed approach combines various techniques, including tokenization, POS tagging, and dependency parsing, to improve the accuracy of SA models. This study not only establishes the effectiveness of transformer architectures in sentiment analysis but also helps the advancement of brand valuation approaches. The findings have impacts on marketers, as they provide a powerful tool for assessing customer sentiment and obtaining brand value with unprecedented accuracy. The results of the proposed model outperform with 95.98% of accuracy, 96.72 % of precision, 94.38% of recall, and 95.53% of F-measure.

Enhancing Machine Learning Resilience to Adversarial Attacks through Bit Plane Slicing Optimized by Genetic Algorithms

2024-06-21T07:49:24+00:00

This research delves into enhancing the resilience of machine learning models, particularly image classification algorithms, against adversarial attacks. The focus is on using genetic algorithms to optimize bit plane slicing configurations, thereby improving the models’ robustness. The study reveals that models with 5-bit depth representations exhibit superior resilience, achieving high accuracies of against FGSM attacks and against DeepFool attacks. These results underscore the importance of adjusting detail levels through bit plane slicing to maintain algorithmic integrity under adversarial conditions. Despite a significant drop in performance due to adversarial modifications, with accuracy falling from to , a notable recovery was observed, highlighting the effectiveness of the optimized defense strategies. The findings advocate for further research into dynamic bit plane slicing and the development of advanced defense mechanisms using genetic algorithms, aiming to bolster the security and reliability of machine learning models against the continuously evolving adversarial threats.

tain algorithmic integrity under adversarial conditions. Despite a significant drop in performance due to adversarial modifications, with accuracy falling from to , a notable recovery was observed, highlighting the effectiveness of the optimized defense strategies. The findings advocate for further research into dynamic bit plane slicing and the development of advanced defense mechanisms using genetic algorithms, aiming to bolster the security and reliability of machine learning models against the continuously evolving adversarial threats.

A Hybrid Recommender System Exploits Layered Convolutional Neural Networks.

2024-06-21T07:52:24+00:00

Enhancing user engagement and satisfaction in e-commerce platforms by incorporating customer preferences and interests into product recommendations is of paramount importance. However, accurately capturing these preferences, both explicit and implicit, from the vast array of products available in catalogues poses a significant challenge. In this study, we propose a novel approach that leverages deep learning techniques to extract latent preferences from product images. Our method focuses on extracting relevant data from the features of interest in product images, thereby enabling the identification of underlying customer preferences. We demonstrate the efficacy of our strategy by integrating this data extraction process into a product-based recommendation algorithm. Through experimental validation, we showcase the effectiveness of our approach in generating personalized suggestions tailored to individual customer preferences. Our findings underscore the potential of deep learning-based methodologies in harnessing visual cues to enhance the personalization of e-commerce recommendations, thereby improving user experience and engagement.

Resource Offloading (Balancing) in Cloud Environment using Particle Swarm Optimization and Improved Particle Swarm Optimization on Xen Server

2024-06-21T07:57:12+00:00

Cloud has numerous strong servers to allot immense solicitation of clients. Load Balancing (Resource Offloading) is a strategy to disperse tasks on numerous VM’s of Server to accomplish Asset usage, Reduce Reaction (cost) time and keep away from trouble. Resource offloading is a crucial factor in ensuring that the available resources are utilized efficiently, and the workload is distributed optimally. This paper presents novel research of resource offloading in a cloud environment and explores the different approaches to resource offloading, including Resource offloading, and Response time reduction of VM. In this article, A PSO & Improved PSO, has been introduced to track down the better arrangement for the issue of Allotment Resources (Assets) and load adjusting in CC. This Work was probed Xen Server and aftereffect of the Proposed Calculation Further developed PSO was extremely uplifting. Altogether the aftereffect of IPSO contrasted and PSO Calculation.

Analyzing the Impact of Lexicon Based Features for Emotion Classification

2024-06-21T07:59:14+00:00

Emotions are psychological states that are frequently represented through actions, words or text. Emotion analysis is a method for deciphering a text to identify the feelings conveyed within it. Identification of emotion(s) contained in music lyrics is a complex process. The emotion model plays a key role in the design of emotion identification algorithms. Several text features are defined and used with machine learning algorithms for labelling lyrics based on emotion. Most of these features are defined following natural language processing concepts. Emotion lexicons play an important role in mapping words that appear in lyrics with discrete and continuous emotions. In this work, we analyze the impact of features derived from lexicons in identifying the underlying emotion of lyrics. Experiments are carried out with emotion-annotated datasets and different lexicons. Classification models are built with the lexicon features. The results obtained highlight the impact of Lexicon based features on classification accuracy. For the design of robust and efficient emotion classifier, the lexicon features need to be combined with other text based features.

Integrating User Attribute Influences and DL-Based Anonymization for Enhanced Privacy Protection in Medical Record Data Sharing for Publishing

2024-06-21T08:01:23+00:00

In today's data-driven healthcare research landscape, sharing medical records for research is pivotal for advancing medical knowledge and patient care. However, ensuring individuals' privacy while maintaining data utility poses a significant challenge. To tackle this issue, this study proposes a novel Attribute Influence Anonymization using RVAE (AIARVAE) for enhancing both privacy and utility in medical records data sharing. The proposed model employs a preprocessing step to identify and filter Quasi-Identifiers (QIs) and Sensitive Attributes (SAs) from the dataset. Then quantify the susceptibility of QIs and measure the uncertainty of SAs using entropy. These metrics are then fed into a Recurrent Variational Auto-Encoder (RVAE) model, which replaces low-entropy SAs with sanitized values with the help of QI values. This approach mitigates the risk of explicit disclosure of private information while preserving data utility. By integrating attribute influences, the proposed model provides a comprehensive solution for safeguarding medical records data privacy during research sharing and promoting responsible and ethical data-driven healthcare research.

Bayesian Analysis of Micro-Expressions: A Study on CASME II and AffectNet

2024-06-21T08:03:31+00:00

This paper provides a thorough investigation into utilizing a Bayesian framework to identify facial micro-expressions. The study uses two separate datasets, CASME II and AffectNet. CASME II is well-known for its high-quality videos that are specifically created to capture subtle micro-expressions in controlled settings, whereas AffectNet offers a wide range of facial expressions captured in more realistic environments. Our research utilizes sophisticated probabilistic models to improve the identification and categorization of brief facial expressions that frequently signify underlying emotions. Our objective is to tackle the difficulties presented by the nuanced and swift characteristics of micro-expressions through the utilization of Bayesian inference techniques. This study showcases the efficacy of Bayesian models in recognizing micro-expressions and emphasizes the significance of dataset characteristics in developing resilient recognition systems. The results promote additional investigation into adaptive models capable of flexibly adapting to the variability in real-world data, potentially resulting in more precise and widely applicable emotion recognition systems. The software used for conducting the experiments is Python.

Hyperparameter Tuning of the LSTM model for Stock Price Prediction

2024-06-21T08:05:31+00:00

The stock market serves as a mirror, revealing the actual state of the nation's economy. Experts can monitor the nation's economic status by following the stock market's fluctuations. Predicting the stock market is so essential in the cutthroat world of today. Because stock prices are chaotic, dynamic, and nonlinear, predicting them is challenging. Stock price forecasting is aided by deep learning methods such as LSTM. The algorithm's forecast is erroneous since its hyperparameter was not chosen correctly.

This work contributes to developing a hyper-tuned LSTM model for stock price prediction. Numerous hyperparameters are included, such as neurons, batch size, epoch, learning rate, and dropout rate. The primary goal is to identify the optimal set of parameters that will enable the LSTM forecasting algorithm to operate at a high level of performance. Three widely used error metrics are used to assess algorithm performance: R², which indicates how well our predictions match the actual data; MSE, which displays the discrepancy between the predicted and actual data; and MAE, which indicates the average deviation between our predictions and the actual data. For training, testing, and validating the data set, values of three error metrics for various parameter combinations are gathered. The best value of these error metrics helps in selecting the best possible combination of parameters.

A proven prediction method called Adaboost is compared with the output error metrics of the LSTM model to confirm our hyperparameter tunning efforts. The LSTM model's potential for precise stock price prediction is strongly confirmed if its error metrics value is comparable to or superior to Adaboost's.

Traffic Prediction Based on Air Quality in IoT-Based Smart City Using Regression and Ensemble Techniques: Bagging and Stacking

2024-06-21T08:07:30+00:00

Traffic forecast implies determining the volume and thickness of the traffic stream, typically for the reason of controlling vehicle development, decreasing congestion, and producing the ideal routes with the least amount of time or energy consumed. Accurate street traffic flow determination is among the foremost essential factors in smart cities. In this research, we utilized air quality data and ensemble regression methods to establish a predictive model for traffic patterns, recognizing the correlation between air pollution levels and congested traffic conditions. This study was conducted in two distinct stages. In the first phase, we compared the performance of 10 different regression models (Decision Tree, KNN, Cat Boost, Linear Regression, Lasso, Elastic Net, Kernel Ridge, Gradient Boost, XGB, and LGBM), and K-Nearest Neighbour gave the best result with RMSE 2.80 and Lasso gave the least performance with 5.28 RMSE. In the second phase, we developed models based on ensemble techniques: bagging and stacking. Depending on the performance of the regressors in the first phase, we attempted numerous permutations of distinctive models in bagging and stacking till we got the most excellent conceivable results. Finally, out of many arrangements, the Stacking Model with CatBoost, KNN, and Decision Tree as base learners and Lasso as meta learner performed better than KNN and Bagging Ensemble Regression models with RMSE 2.09.

Enhancing Predictive Accuracy in Phishing Attack Detection: A Study on the Impact of Collinearity and Feature Selection in ML-based Logistic Regression Models

2024-06-21T08:10:23+00:00

Phishing threats present dangers, for people and businesses alike emphasizing the need, for creating reliable detection techniques. It is crucial to establish phishing tactics to protect confidential data and avoid monetary damages. This study delves deeper into the intricacies of logistic regression models and how these models could effectively detect phishing attacks with a focus on impact of factors like collinearity and feature selection on predictive accuracy and model performance. In addition to logistic regression, different machine learning models, such as Decision Tree Classifier, Gaussian Naive Bayes, Logistic Regression, K Nearest Neighbors and Linear Discriminant Analysis were also considered to analyze the relationships between predictor variables and successful phishing attack likelihood and the predictive accuracy from each of the methods. By conducting experiments and comparisons we show that addressing collinearity issues and employing feature selection techniques significantly improve the predictive accuracy of logistic regression models compared to other common machine learning models. Through a methodical process of feature engineering focused on addressing collinearity among predictors, we achieved a substantial reduction of over 35% in the false negative rate for the logistic regression model which is crucial as false negatives are more costly. These findings provide insights, for enhancing the efficiency of phishing detection systems to strengthen cybersecurity defenses against emerging threats.

Environmental Regulations in Computing: Policies and Practices

2024-06-21T08:12:36+00:00

In an era where environmental concerns are vital and people are fighting to save environment, Green Information and communication Technology appears as a miraculous boon for the World. Green ICT talks about uses of technology in environment friendly way. Information and communication technology (ICT) plays continuously a vital role in encouraging sustainability. Green ICT, or Green Information and Communication Technology, is creating, utilizing, and disposing of ICT resources in an environmentally sustainable manner. This includes lowering energy use, eliminating electronic waste, and leveraging technology to encourage sustainable behaviours across multiple sectors. The government has made various policies for the use of ICT in environment friendly way. This paper examines the Environmental policies and practices that are implemented to ensure the protection and preservation of the environment in relation to computing activities..

Recommendation System using Neutrosophic Logic in Agriculture

2024-06-21T08:15:00+00:00

For billions of people worldwide, agriculture provides food and a means of subsistence, making it an essential component of the global economy. Agriculture, one of the main industries, contributes to both economic stability and food security. To fulfill the expanding nutritional needs and maintain long-term resilience, however, innovation and sustainable techniques in agriculture are urgently needed as a result of the COVID-19 pandemic and continued climate change. Computational Intelligence is becoming more and more applicable in several automobile, industrial, and commercial sectors worldwide. Its ability to provide efficient and accurate functionalities attracts top companies to invest in A.I. because scientists and researchers believe that it will have a significant implication in the strife towards improving human life. Cultivating crops unsuitable to environmental conditions, such as soil and weather, is one of the main reasons behind the continuing decline in agricultural advances. One way to solve this problem is to apply the use of a recommendation system to predict favorable crops. Here, we are proposing a recommendation system based on neutrosophic logic. Neutrosophic logic is a promising tool for smart agriculture that can cope with the complexity and dynamism of agricultural systems. By incorporating neutrosophic logic into smart agriculture via IoT, it is possible to achieve more accurate, reliable, and robust solutions that can improve the quality and quantity of agricultural outputs while reducing the environmental and social impacts. The proposed model efficiently predicts the crop yield outperforming existing models like KNN, fuzzy logic, and neutrosophic logic.

Analyzing the Effect of Different Activation Functions in Deep Learning on Accuracy and Execution time

2024-06-21T10:25:30+00:00

Activation functions is critical in specifying the active node within neural networks. Choosing the most suitable activation function is crucial because to its impact on the overall output of the network. Prior to choosing an activation function, it is essential to check the characteristics of each activation function based on our specific needs. The monotonicity, derivatives and range of the activation function are important characteristics. In our review study, we examined 13 different activation functions, such as ReLU, Linear, Exponential Linear Unit, Gaussian Error Linear Unit, Sigmoid, SoftPlus, among others.

Building Trust on the IoT Connected World: Addressing Security Challenges in IoT Architectures and Applications

2024-06-21T10:27:38+00:00

IoT is revolutionizing how we interact with the world, connecting everyday objects and enabling a vast array of applications. However, this interconnectedness raises critical security concerns. This paper delves into the foundation of the IoT, exploring various architectures that support its functionality. We then examine the diverse applications that leverage these architectures, highlighting the potential benefits they offer across various domains. However, the paper argues that without robust security measures, the true potential of IoT cannot be fully realized. We analyse the vulnerabilities inherent in IoT systems, exploring common security issues such as weak authentication, data breaches, and botnet attacks. To address these challenges, the paper investigates existing and emerging solutions that can fortify the security posture of the IoT ecosystem. This includes exploring secure communication protocols, encryption techniques, and leveraging advancements in technologies like blockchain and machine learning. By providing a comprehensive understanding of IoT architectures, applications, and security considerations, this paper aims to guide researchers and developers in building a more secure and trustworthy foundation for the future of the IoT.

System Model and Problem Formulation to Address Performance Issues in Edge Intelligence

2024-06-21T10:30:17+00:00

When data processing is implemented in close proximity to end devices with intelligence and ample capabilities, it not only improves real time processing but also increases the effectiveness of generated results and reduces a significant burden on the overall network. Various metrics, such as computational speed, reaction time, CPU demand, network demand, and delay sensitivity, play a crucial role in enabling edge devices to execute complex tasks within time constraints. This paper presents an approach by adopting fuzzy logic to transmit the incoming tasks from the edge devices to one of the edge-cloud servers, which is decided by the edge orchestrator, taking into account various application characteristics. The primary aim of the proposed approach is to enhance task offloading by reducing service time and boosting the efficiency of edge devices. A system model and problem formulation have been designed with the help of which QoS parameters are improved in an edge-cloud environment by taking into consideration the balancing workload among the resources in the network.

A Secure E-voting System based on Elliptic Curve Digital Signature Algorithm with Hybrid Consensus Mechanism

2024-06-24T10:46:16+00:00

Background: The evolution of technology has brought about significant changes in many existing processes, making them simpler and safer. Electronic voting (E-voting) is a notable example that has replaced traditional voting systems to achieve accurate and reliable results with minimal human interference. However, E-voting faces significant challenges such as vote rigging, vote theft, and various other security threats. Methods Used: To address these security concerns, the elliptic curve digital signature algorithm with hybrid consensus algorithm (ECDSA-HCA) was employed. A secure web-based E-voting approach has been developed to facilitate end-to-end communication between users, ensuring the prevention of vote theft during the polling announcement in the nation. The ECDSA-HCA involves three main stages: the registration process, polling, and the announcement of results. In the proposed model, the election commission utilizes blockchain technology to verify and validate vote data. Subsequently, the ECDSA-HCA method is employed to securely store voter data in the blockchain, utilizing encryption and an e-voting cloud system (ECS) tailored to the data structure of user-specific modelling processes. Results achieved: Upon analyzing the results, it becomes evident that the proposed ECDSA-HCA approach outperforms in terms of communication time (1871 μs), encryption time (1650 μs), latency (24 ms), and throughput (63 Tps). Concluding remarks: In this study, the number of users is extended to 1000 by conducting a simulation network five times, each with 200 users as the size of each set of nodes. To assess the effectiveness of the ECDSA-HCA, it is compared to existing studies such as ECS and ECDSA.

Improving Opinion Mining Accuracy with Dragonfly Algorithm-Based Hybrid Classification

2024-06-24T10:48:52+00:00

The paper proposes the Dragonfly + Hybrid Classifier, a novel approach designed to enhance opinion mining across diverse datasets. Leveraging the Dragonfly algorithm for feature set selection and combining it with a hybrid classification method, this innovative approach offers the potential for more accurate and reliable predictions. On the Twitter Sentiment dataset, notorious for its dynamic and noisy nature, the Dragonfly + Hybrid Classifier excels with an average precision of approximately 0.93498, recall of approximately 0.92965, and an F-measure of approximately 0.93208, alongside an average accuracy of around 96.134%. Within the Movie Review dataset, where opinions are nuanced and context-dependent, the Dragonfly + Hybrid Classifier secures an impressive average precision of approximately 0.91348, coupled with an average recall of approximately 0.9189, achieving an F-measure of approximately 0.91582 and maintaining an average accuracy of around 94.98%. In the context of the Depression dataset, where sensitivity and accuracy are paramount, the Dragonfly + Hybrid Classifier excels with an average precision of approximately 0.9627, an average recall of approximately 0.966, an F-measure of approximately 0.9643, and an average accuracy of around 94.62%. These findings collectively affirm the Dragonfly + Hybrid Classifier as a potent tool for opinion analysis across diverse domains, positioning it as a valuable asset in field of opinion mining and analysis applications, particularly in domains where opinion understanding is paramount.

An Artificial Intelligence Based Optimal Technique for Coordinating Multiple Setting Group Type Overcurrent Relays Considering Network with Renewable-Resources

2024-06-25T07:45:53+00:00

Integrating renewable based resources with the network changes the power flow and makes relay coordination more complex and challenging. Looking at the supportive environment for renewable based distributed generation on the grid, the penetration of RE-based distributed generation continuously increases. Relay coordination optimization challenges are complex problems that involve several objectives and constraints to DGs. The aim is to minimize the operation time of each relay while ensuring that all relevant constraints are met. The coordination constraints consist of fulfilling the coordination time interval (CTI) and adhering to the time requirements for relay operations. In order to address coordination optimization problems, decision-making variables are established to minimize the objective function while ensuring that the essential constraints are satisfied. Most available research focuses on defining relay coordination for fixed network topologies. However, relay coordination varies in practice due to many factors, such as maintenance and element failures. Modern overcurrent relays can store multiple relay settings, but the number of settings they can store is limited compared to the vast network configurations. To address this, a suitable clustering technique like K-Means can be employed to group different network topologies using an appropriate clustering index. The target function and the constraints are both time-dependent, so the standard index fault current deviation is not used as the clustering index. Instead, time-dependent features like the average relay operational time are used. IEEE-14 bus system along with wind and solar type DG has been used to address the optimization problem.

Rest-Rec: Restaurant Recommender System Based on Model Based Collaborative Filtering Approach

2024-06-25T07:48:16+00:00

Recommendation Engine has become the need for everyone and has changed the lifestyle of people in the aspect of searching products and services. Recommendation systems are used in almost all areas driving from education to entertainment. A recommendation system is a class of information filtering system to provide choices to the users based on their preference. Considering the need and importance of recommendation system, this paper proposed a recommender called Rest-Rec for restaurants based on collaborative filtering approach. Rest-Rec analyses the previous user’s information and recommends the restaurants as per user’s preference. K-Means algorithm is employed to cluster the restaurants based on the rating by the users. Performance of proposed Rest-Rec is evaluated using data from Trip Advisor website in terms of Precision, Recall, F1-Score. It is evident from the results that Rest-Rec provides recommendation with precision of 95.67%.

K-Means Segmentation and Normalized Histogram: An Effective Method for Detecting Brain Tumor from MRIs

2024-06-25T07:50:15+00:00

In the medical field, brain tumors are evaluated using a technique known as magnetic resonance imaging or MRI. To eliminate the additive noise which is found in MRI images, including Gaussian, Salt & Pepper, and Speckle noise, this study employs a technique to examine and classifies image d-noising filters including Mean, Adaptive, Minimizing, UN-sharp masking filter and the Gaussian filters. The DE-noising efficiency of each method is investigated using PSNR and MSE. The effective brain tumor segmentation utilizing the normalized histogram and the K-means clustering algorithm is shown as a novel method. Support Vector Machine (SVM) is to provide accurate forecasting and classification.

Flexible and Cost-Effective Spherical to Cartesian Coordinate Conversion Using 3-D CORDIC Algorithm on FPGA

2024-06-25T08:16:24+00:00

In computer science, transforming spherical coordinates into Cartesian coordinates is an important mathematical operation. The CORDIC (Coordinate Rotation Digital Computer) iterative algorithm can perform this operation, as well as trigonometric functions and vector rotations, using only simple arithmetic operations like addition, subtraction, and bit-shifting. This research paper presents hardware architecture for a 3-D CORDIC processor using Quartus II 7.1 ALTERA software, which enables easy modifications and design changes due to its regularity and simplicity. The proposed 3-D CORDIC model is evaluated by comparing the calculated results with the simulated results to determine its accuracy. The results were satisfaction and the proposed model could be suitable for numerous real-time applications.

A Unified Approach for Network Intrusion Detection using Ensemble Machine Learning Classifier using Support Vector Machine and Naive Bayes

2024-06-25T08:39:56+00:00

The Internet and communication areas are developing at a rapid pace, which has increased network size and data demand. Consequently, this surge has given rise to numerous new attacks, posing significant challenges for network security, which are notoriously difficult to pinpoint accurately. Reviewing existing literature reveals that intruders employ sophisticated intelligence and tactics to create these threats, making their monitoring and detection quite challenging. This underscores the critical importance of network data security over the open web. Hence, it becomes imperative to develop a security mechanism that can effectively monitor network traffic to identify and detect these threats.One such potent security measure discovered to tackle these challenges is an Intrusion Detection System (IDS). Many IDS techniques leverage various Machine Learning (ML) algorithms to safeguard data against a range of network attacks. In the past, ML methods have typically centered on creating a solitary model for intrusion detection. Yet, it is widely acknowledged that no individual machine learning algorithm can effectively manage all forms of network attacks. Hence, this study primarily emphasizes the proposal of an ensemble classifier merging the strengths of the Support Vector Machine (SVM) and Naive Bayes (NB) algorithms. This strategy aims to bolster the efficiency of network intrusion detection through meticulous monitoring of network traffic data.

Enhancing Agile Software Development: A Comprehensive Framework for Metrics-Driven Performance Evaluation

2024-06-25T08:43:49+00:00

In the realm of Agile software development, the quest for efficient performance evaluation methodologies remains paramount. Grounded in empirical research and industry best practices, our framework offers a systematic approach to gauge the efficacy and productivity of Agile teams through a meticulous selection of metrics. Emphasizing the significance of quantitative analysis, our framework advocates for a balanced blend of traditional and Agile-specific metrics, encompassing aspects such as velocity, cycle time, and customer satisfaction. By leveraging this comprehensive array of metrics, organizations can gain nuanced insights into team dynamics, project progress, and overall performance, thereby fostering a culture of continuous improvement and informed decision-making. Furthermore, our framework incorporates mechanisms for adaptability, acknowledging the dynamic nature of Agile environments and the need for iterative refinement. Through a rigorous validation process involving real-world case studies and industry feedback, we demonstrate the practical applicability and efficacy of our framework across diverse Agile contexts. Ultimately, our research contributes to the advancement of Agile software development practices by providing a robust foundation for objective performance evaluation, facilitating the pursuit of excellence and agility in software delivery.

An Improved Gazelle Optimization Algorithm for Influence Maximization to Identify Influential Nodes in Social Networks

2024-06-27T08:55:49+00:00

The goal of the Influence Maximization (IM) issue is to choose a component of the k-most influential nodes in a system so that the amount of influence spread by the seed set is maximized.When the transmission probability is high, greedy algorithms have a difficult time effectively approximating the predicted spread of influence of a particular node set and are not readily scalable to large-scale systems.Low solution accuracy or high memory costs are common issues with traditional heuristics based on constrained diffusion channelsor network topology. To address the IM issue more effectively, an Improved Gazelle-Based Optimization Algorithm for Influence Maximization (IGOA-IM) is proposed in this research.A unique local exploitation technique that combines random walk and deterministic procedures is proposed to enhance the suboptimal meme of everymemeplex to facilitate the global exploratory solution.The study findings on the spread of influence in twelve real-world networks demonstrate that IGOA-IM outperforms numerous state-of-the-art alternativesfor IMin choosing targeted influential seed nodes.

Identification and Categorization of SMS using Deep Learning and Machine Learning Methods

2024-06-27T08:57:56+00:00

Text messages are short messages that can be used for personal as well as professional ways to share messages without involving the internet as a mode of communication. There are some essential text messages and some are nonessential. It is crucial to filter out the nonessential messages from the essential ones. Various machine learning and deep learning methods are used to categorize the text messages. This research work uses various machine learning and deep learning methods to categorize them. To extract the features from the text messages this study uses word embedding and contextual embedding techniques. Finally, the measurement of the performances is done with the help of performance matrices and confusion matrix parameters. For the word embedding-based feature selection method the Extra Tree and LSTM are more accurate i.e. 96.86% and 98.06%. And for the sentence embedding-based feature selection method the SVM and Bi-directional LSTM are more accurate i.e. 99.1% and 99.19%.

Blockchain Adoption in Real Time Applications: Characteristics, Challenges and Prospects

2024-06-27T08:59:58+00:00

This paper investigates the different characteristics of blockchain which helps real time applications to overcome the challenges they are currently facing. A detailed study of adoption of blockchain in real time applications is performed. To achieve the aim of this study, the authors systematically reviewed the literature in order to provide answers to challenging queries regarding blockchain technology. We created a python function that looked through numerous databases to find blockchain adoption research work in different real time applications. We observed that healthcare, education, Internet of Things, finance, identity management, and government services are the major areas where blockchain is gaining a lot of attention. In addition to this, the impact of widely used blockchain features in these areas are also identified. In this study we provide an approach to identify potential barriers in adoption of blockchain in real time applications using the Decision-Making Trial and Evaluation Laboratory (DEMATEL) tool. The objective of this approach is to prioritize these barriers based on their degree of influence while simultaneously examining and illuminating the causal relationships between them.

Technical and Economic Analysis of PV Integrated DC and AC in Front of the Meter System for Automated and Manual Dispatch Modes

2024-06-27T09:01:47+00:00

Front-of-the-meter (FOM) based PV system is helpful in adding flexibility to distribution networks for locally generating renewable energy to decarbonise the environment from the impacts of power generation sector. Presently FOM systems are used for specific services at distribution level but for applications in industry these are under development. In this paper a techno-economic analysis of front-of-the-meter systems in primary networks with PV system is presented that covers (i) impact of battery storage systems types (ii) the in terms of different dispatch scheme; and (iii) the quantification for AC and DC connected schemes that give profitability based on techno economic analysis point of view. The analysis performed for grid-level battery energy storage technology particularized to weather data and electric tariff rates followed in Lucknow U.P,India. The techno-economic analysis is covering one year time duration with 30 minute resolution executed at System Advisor Model (SAM) tool. The results are showing that, the techno economic benefits of FOM systems are better than traditional transmission-level services. Several approaches are observed for improving systems profitability.

Data-Driven Intelligent Clustering-Based Optimization for Enhancing Urban Logistics Delivery Systems: A Case Study in Casablanca, Morocco

2024-06-27T09:04:00+00:00

These In the field of supply chain management, making sure products get to people efficiently is not easy. There are issues like customers being unhappy, figuring out the optimizing itinerary of trucks, how much they can carry, and optimizing the delivery time. In this paper, we are introducing a smart system that makes the whole delivery process smoother, starting from managing inventory to reaching the customers. Our system leverages clustering techniques to automate and simplify this complex process. We collected data and tested our approach on a mass retail company in Casablanca, Morocco. This data includes information about customer locations, order details, and the available delivery trucks with their capacities. At the core of our solution lies a unique clustering algorithm, custom-made to handle our specific challenges. The approach starts by defining how far apart cus-tomers' locations can be, ensuring we don't group locations that are too distant from each other. Then, we use a straightforward method to create these groups based on proximity between locations and order details. This ensures the efficient allocation of customer orders to clusters, maximizing truck fill rates. In short, our innovative approach streamlines delivery operations, reduces cus-tomer complaints, optimizes fleet management and guarantees on-time, cost-effective deliveries with a highly satisfactory service rate.

Bee vs Wasp Classification Using Advanced Deep Learning Techniques: CNN, VGG 16

2024-06-27T09:06:12+00:00

This paper explores the use of Convolutional Neural Networks (CNNs) and ResNet-34 architecture for grasshopper and grasshopper classification and discrimination from image datasets Using the deep learning capabilities of CNNs and the rest of ResNet-34 learning a, we address image recognition challenges in biological monitoring. Various data sets of bee and wasp images were used to train and validate the ResNet-34 model. The model performed better than traditional methods and achieved high accuracy in discriminating between two groups of insects. This study demonstrates the potential of CNN and ResNet-34 to automatically identify insects, supporting biodiversity research and conservation efforts.

Adaptive Threshold-based Reserved G2NN Feature Matching with Hybrid Deep Feature Learning for Copy-Move Image Forgery Detection

2024-06-27T09:08:33+00:00

Different image editing devices have been utilized for performing image forgery activities on social media in recent days. Then, the copied images are placed in various locations of the image. But, the important disadvantage of using these forgery detection approaches is detecting the tampered regions with less efficiency. The ultimate aim of this scheme is to investigate novel copy-move image forgery identification with the assistance of deep learning and matching procedure. In the first step, the benchmark datasets are gathered from different public sources and perform pre-processing using the Weiner filtering and contrast stretching process. Further, the feature extraction is done by a new hybrid deep feature learning method that integrates both the deep learning network called Enhanced Convolutional Neural Network (CNN) and Speeded Up Robust Features (SURF). With these hybrid features, feature matching is accomplished by the improved technique termed Adaptive Threshold-based Reserved Generalized Two Nearest Neighbourhood (G2NN) Feature Matching (AT-RG2NN-FM). The significant intention of the implemented scheme is to perform the optimal feature matching to attain the maximum detection rate. The performance of CNN and feature matching is enhanced by an Intermixed Forest and Cuckoo Search Algorithm (IFCSA). The experimental validation proves the effectiveness of the developed model.

PneuDetect: Pneumonia Detection using a Novel Two-Stage Deep Learning Pipeline from Chest X-Rays – A Review

2024-06-27T09:10:43+00:00

Cognitive Radio Networks (CRNs) play a pivotal role in addressing the spectrum scarcity Challenge by enabling secondary users (SUs) to dynamically access underutilized spectrum bands while ensuring minimal interference with primary users (PUs). In this study, we propose a novel approach that leverages Design of Experiments (DoE) principles to optimize spectrum utilization in CRNs. The research work is carried out to optimize the resources depending on the price factor and the demand in the current scenario. In such case, the demand raises from the secondary users to utilize the frequency spectrum. The primary users take a decision on the design of the experiment. The research work is carried out by designing of experiments. In this research work, Taguchi method, screen design and item factorization are implemented to determine the pricing of the spectrum for utilization by the secondary users with reference to the availability and the prices. The approach ensures efficient utilization of available spectrum resources while maintaining PU protection.

A Fast Simple Linear (FaSL) Unsupervised Feature Extraction Method

2024-06-27T09:20:16+00:00

The increase in volume of high-dimensional data necessitates the use of dimensionality reduction strategies (DRS), which reduce dimensions and extract meaningful insights by eradicating irrelevant features. Linear and nonlinear are the two types in DRS. Nonlinear dimensionality reduction methods have gained considerable popularity in recent years due to their effectiveness in handling real-world datasets with complex nonlinear structures. However, there are some fields where linear data sets are frequently used, including physics, economics, health informatics, social sciences, etc. The major drawback of many existing linear and nonlinear DRS models is their computationally expensive nature. To address this issue, a fast, simple, linear (FaSL) unsupervised feature extraction method is proposed using descriptive statistics. The FaSL performance is evaluated by applying clustering on various benchmark data sets and compared with five linear state-of-the-art methods. The experimental results demonstrate that FaSL outperforms other linear models such as PCA, LDA, LPP, ICA, and FA in terms of accuracy and computation time. The average accuracy improvement of FaSL over PCA, LDA, LPP, ICA, and FA is, in order, 3.4, 9.2, 5.67, 3.97, and 0.075 while reducing computational time by 2.26, 3.1, 1.29, 7.58, and 6.2 times, respectively.

Building Trust in Artificial Intelligence: An Explainable Deep Learning Framework for Brain Disease Detection

2024-06-27T09:23:39+00:00

Artificial Intelligence (AI) has shown promising results across various research fields. However, there is significant concern about its application in medicine due to the critical need for high accuracy and reliable data in this field. A major issue with many existing machine learning models is their lack of transparency; they do not explain the reasoning behind their outputs. This opacity leads to a lack of trust among healthcare professionals, who are hesitant to rely on such technology for critical decisions. Our research aims to address this concern by developing an Explainable Artificial Intelligence (XAI) model. This model not only classifies MRI images but also provides clear explanations for its predictions. By highlighting the specific regions of the brain that influenced each decision, our XAI model helps bridge the gap between AI and clinical practice. This approach empowers clinicians to identify brain diseases more confidently and accurately, fostering greater trust in AI-driven diagnostic tools.

Advanced Diabetic Retinopathy Detection and Classification Using Lightweight Deep Learning Techniques

2024-06-27T09:25:33+00:00

Diabetic Retinopathy (DR) is an ocular disorder that has the potential to result in visual impairment and complete loss of vision in those diagnosed with diabetes. This illness affects the retinal blood vessels inside the light-sensitive tissue layer at the posterior of the eye, known as the retina. This paper presents a complete approach to diagnosing and categorizing diabetic retinopathy using deep learning models. A lightweight Convolutional Neural Network (CNN) is used to detect diabetic retinopathy in fundus images. This CNN has been developed to have fewer parameters and calculations, making it suited for resource-constrained environments while retaining decent performance. The categorization of diabetic retinopathy is carried out with the help of EfficientNet. This model uses an innovative compound scaling approach to strike a balance between the model's depth, width, and resolution. As a result, it maximizes computing efficiency while preserving high accuracy. The proposed detection model obtained an accuracy of 95%, and the classification model produced an accuracy of 84%.

Spectrum Allocation in Cognitive Radio based Traffic Monitoring System Using Machine Learning

2024-06-27T09:27:58+00:00

Vehicle tracking and Traffic Monitoring is essential as it forms the main dimension of a smart city. Globally, during the last decade the number of automobiles in roadways has increased drastically. Traffic monitoring in such a high traffic density era is significantly difficult in various developing countries. Hence, the work focuses on regulating traffic jams by tracking the vehicle and transmitting the data to the regulating authorities in shorter duration with the help of Cognitive Radio technology. The CR technology is very useful for effective traffic monitoring to transmit the traffic management parameters by exploiting Primary User’s (PU) spectrum. For spectrum detection and allocation for high-speed transmission of traffic parameters, various tree related machine learning algorithms like random forest, decision trees and XGBoost are used, examined and compared for better results. Of these, random forest gives high accurate prediction of available spectrum and allocation. On applying the model, we ensure that timely delivery of traffic monitoring information can help in better traffic management and vehicle tracking.

NGBFA Feature Selection Algorithm-based Hybrid Ensemble Classifier to Predict Cervical Cancer

2024-06-27T09:34:45+00:00

Early diagnosis may cure cervical cancer. Researchers have struggled to prediction the disease's course because there are no early indications. Several machine learning methods have predicted CC in the past decade. Ensemble techniques generate and integrate several models for more accurate results. This contrasts with single-classifier prediction. During this research, we established "Robust Model Stacking: A Hybrid Ensemble." This classifier runs a homogeneous classifier-based classifications at the base level, then a heterogeneous ensemble that predicts additional data using majority voting (soft). This study included 858 patients, 32 risk indicator characteristics, and four CC diagnosis test targets. SMOTE oversampling solved the data imbalance problem. For each of the dataset's four goal variables, accuracy, recall, f1-score, precision, and AUC-ROC were used to assess the model. The proposed biopsy approach is 98% accurate, Hinselmann 97%, Schiller 96.09%, and Citology 93%. Ensemble learning improves prediction accuracy and reduces bias and variation in this study.

Electromagnetically Coupled MIMO Antenna for 5G Communication

2024-06-27T09:37:56+00:00

This article introduces a quad-port antenna for 5G wireless communications. The antenna is fed by means of aperture-strips electromagnetic coupling. The optimized slots inside the patch induces strong electromagnetic fields which excites the dielectric resonator (DR). The DR is kept above the slits of the patch for induction of resonant mode. The antenna has bandwidth of 3.53% and 256% at 3.39 GHz and 4.68 GHz. The mechanical dimensions of the antenna are 30 mm x 30 mm. The antenna has full ground which works as a reflector and presents higher peak gain in boresight worth 2.95 dBi and 2.42 dBi. The diversity parameters of MIMO are at par with communication requirements.

A Survey on Load Balancing Algorithms in WSN

2024-06-27T09:39:42+00:00

The key idea in distributed systems and computer networking is load balancing, which divides workloads and incoming network traffic among several servers or resources explores the topic of load balancing in wireless sensor networks (WSNs), emphasizing resource optimization and fair task distribution across sensor nodes. The assessment covers a variety of load balancing techniques, addressing issues with data throughput, network lifetime, and energy consumption. The output sums up a thorough understanding of load balancing methodologies in wireless sensor networks (WSNs), provide guidance for future improvements in network resilience and efficiency. Investigates WSN load balancing strategies with the goal of improving network performance by effectively allocating workload across nodes. It looks at several methods while taking into account things like data accuracy, node lifetime, and energy consumption. An overview of the main load balancing techniques and their uses is given in this study.

Role of Robotic Process Automation and Navigation System in Transport Sector

2024-06-27T09:42:18+00:00

Robotic process automation in transportation can lead to a setting that allows for semi- or completely autonomous navigation. In semi-autonomous mode, the system accepts conventional motion instructions through voice activation or a standard joystick interface and provides robotic movements with obstacle and collision avoidance features. While the fully autonomous mode trials are highly encouraging, the sparsest or semi-automation navigation option is the sparsest or semi-automation navigation mode. Financial savings, higher quality, and better customer experience are just a few of the advantages of robotic process automation (RPA).RPA is a technical application that uses business logic and structured input to automate business tasks. Using RPA technology, a company may develop software, or a "robot," to record and understand applications for processing a transaction, modifying data, triggering responses, and integrating with other digital systems. Businesses may utilize RPA to automate mundane rules-based business procedures, freeing up business users’ time to focus on customer service or other higher-value duties. It is crucial to the advancement of public transport. However, the company is currently experiencing major difficulties. The primary goals of this research are to assess the operational and financial performance, as well as the function of robotic navigation systems in the transportation industry. This research uses both primary and secondary data. It contains interviews with different metrics for data analysis that include operational and financial characteristics such as fleet, collection, and passengers, among others. A navigation system is a two-way communication system, similar to a digital telephone, that connects with a central service center to ascertain the user’s actual location and navigational information. Ideally, a human-to-human information interface is provided. Some of the components can be stored at a remote and fixed base station due to the usage of a two-way communication system.

Exploring the Landscape of Communication Technologies and Networking: An In-Depth Analysis

2024-06-27T09:48:36+00:00

At this point in time, networking and communication technologies are very necessary in order to facilitate the seamless connection of individuals and the exchange of information. Several different kinds and categories of computer networks may be found within the complex web of computer networks, each of which serves a different function. In this book, the principles of effective data transmission and communication via networks are dissected and discussed. The data that was gathered and statistically evaluated came from primary sources, which included in-person interviews. The purpose of this was to highlight the value of networks in our fast-paced world. The need of having an appropriate network design is shown by a graphical representation of computer networks as well as a study of the fundamental components that comprise them. It is possible that companies and researchers may profit from a network that is well-structured since it could increase their processing capacity and overall productivity. In order to effectively navigate the intricate world of communication technologies and achieve ideal network performance, it is necessary to have a harmonic balance of the appropriate network components, intelligent design, and professional IT implementation. This book provides a comprehensive introduction of the complex world of networks, highlighting the important role that networks play in promoting connectedness and furthering technological and process innovation.

Machine Learning-Based Gait Analysis for Distinguishing Older and Younger Walking Patterns in Neurodegenerative Diseases

2024-06-27T09:50:38+00:00

Neurodegenerative diseases like Parkinson’s Diseases, Alzheimer’s Diseases, Multiple sclerosis and Huntington’s disease can severe a person’s walking style due to their impact on the brain and the nervous system. Gait analysis, which involves the study of a person's walking pattern and movement, plays a crucial role in the diagnosis and monitoring of these diseases. By examining changes in gait parameters such as stride length, walking speed, and balance, healthcare professionals can gather important information about the underlying neurological impairments and track disease progression. Gait analysis involves the measurement of various parameters, including the stride interval. Changes in the stride interval can indicate alterations in motor control and gait stability, allowing healthcare professionals to assess the severity of neurodegenerative diseases and monitor the effectiveness of treatment interventions. There is lack of research in studying the effect of Continuous Wavelet Transform (CWT) in stride intervals of the young people and old people. It is not clear whether the CWT is a feasible feature extraction method to classify the stride interval of old people and young people. The objective of this paper is to apply Support Vector Machine (SVM), K-Nearest Neighbors (KNN) and Random Forest algorithms to the maximum Root Mean Square (RMS) value of CWT to determine the most effective machine learning techniques for distinguishing between older and younger walking patterns. KNN stands out the best in performance by scoring 93% for all weighted average (precision), weighted average (recall) and weighted average (f1-score). SVM comes out second in performance by scoring 86% for weighted average (precision), 83% for weighted average (recall) and 84% for weighted average (f1-score) with the shortest processing time, 3.2302s. From the boxplot of the Maximum CWT RMS of the young and the old people, it can be seen that the stride interval of the young people is higher and more diverse than the old people.

Enhancing Efficiency and Performance with RFT Gate Designs

2024-06-27T09:52:59+00:00

Ensuring congruence between the input vector's parity and that of the output vector stands as a pivotal requirement within fault-tolerant reversible logic gate circuits. This parity-conserving attribute assumes paramount significance in discerning latent anomalies within the circuitry. Particularly in the realm of nanotechnology applications, this trait serves as a linchpin in crafting fault-tolerant systems. The realm of reversible logic gates (RLGs) designated for fault tolerance enjoys widespread recognition for its adeptness in nascent computing paradigms, exemplified by optical and quantum computing. Within our discourse, we introduce the concept of reversible fault-tolerant lookup tables (LUTs), wielding profound influence in the evolution of Configurable Logic Blocks (CLBs). This architectural construct of CLBs amalgamates fault-tolerant reversible logic constituents, encompassing D-Latches, Master-Slave Flip-flops, and Multiplexers. Our proposed architecture undergoes exhaustive simulation, logical scrutiny, and FPGA Spartan 3-based implementation. The resultant simulation outcomes and practical deployments corroborate the efficacy underlying the design of reversible fault-tolerant (RFT) gates. Particularly noteworthy are the discernible reductions observed in power dissipation and latency within these gates. In the precincts of 90 nm CLB technology, a conspicuous power decrease of around 95.5% is discerned, accentuated further to 98% in the context of 45 nm CLB technology. These empirical revelations serve to underscore the inherent potential of reversible fault-tolerant gate configurations, poised to augment efficiency and performance benchmarks across FPGA applications.

Integrating Algorithms with Intuitive AI to Forecast the Likelihood of Cerebral Infarction in Patients Exhibiting Signs of Illness

2024-06-27T10:07:01+00:00

Stroke is a highly debilitating disease that is widespread globally. It is a major public health concern that requires urgent attention. Throughout their lifetimes, individuals and their families may experience the severe consequences of this complex and diverse neurological disorder. These consequences can be encountered by individuals. This case study examines the intricacies of stroke, encompassing its etiology, potential risks, manifestations, diagnosis, and therapeutic interventions using Intricate Artificial algorithm to forcast and predict the occurances of stoke using available patient symptoms. The system uses cluster grouping and random forest model to accurately predict the occurance of stroke based on lifestyle and symptoms of a group of patients classified based on gender It also encompasses concerns over the potential hazards linked to stroke. Moreover, the entire narrative underscores the importance of immediate action and comprehensive medical intervention. If there is a sudden interruption of blood flow to the brain, a stroke, often known as a "brain attack," will occur instantly. Consequently, the brain cells will be deprived of the necessary oxygen and nutrients required for optimal functioning. This interruption, which can be caused by clots (ischemic stroke) or ruptured blood vessels (hemorrhagic stroke), has the ability to cause damage to the neurological system and, in the most severe situation, permanent disability of the affected individual. Due to the significant impact of stroke on individuals' everyday functioning and quality of life, research on stroke is highly crucial.

Comparative study of ΔE on Different Label Stock on Digital Dry Toner Electrophotography and Inkjet Digital Printing Technique

2024-06-28T06:59:40+00:00

The printing and packaging industries grow fastest across the world, as well as in India. Printing labels have made a major contribution to the growth of the printing and packaging industries. In the printing industry, labels are printed using different printing techniques. In the modern era, digital printing has rapidly grown across the world. In this research, comparisons were made between different label stocks printed by dry-toner electrophotography and inkjet printing technology. After the data analysis, the print quality of chromo paper was fine as compared to PP white paper in the case of chromo paper and PP white paper printed by dry-toner electrophotography. The print quality of PP white paper is better as compared to chromo paper in the case of both substrates printed by inkjet printing technology.

Enhanced Autism Severity Prediction: Hybrid Gradient Boosted Tree and Deep Learning Models

2024-07-01T08:50:53+00:00

Autism Spectrum Disorder (ASD) is a condition of developmental disability impacting both behavior and brain functionality. It cannot be diagnosed through medical tests; hence, the diagnosis relies heavily on historical data. Data science models, like Gradient Boosted Trees and Deep Learning, play a crucial role in predicting autism risk by evaluating relevant information and identifying patterns. This paper proposes a novel Hybrid Model that combines the advantages of both Gradient Boosted Tree and Deep Learning models. The aim is to reduce the number of necessary diagnostic tests for autism, thereby offering potential solutions for the healthcare sector. This model achieved an accuracy of 95.52% in predicting the severity of autism using historical adult autism data. The historical patient data used for this study is available on the Kaggle Repository. This perspective highlights the crucial importance of data science in diagnosing healthcare issues.

Artificial Intelligence-Driven Forensic Analysis of Digital Images for Cybersecurity Investigations

2024-07-02T09:24:42+00:00

The current world is highly circumscribed by these numerous threats hence providing a basis for developing better ways of handling images for cybersecurity purposes specifically in the field of cyber forensics. This paper aims to compare and analyse how AI can be implemented to increase the effectiveness and timeframe of forensic analysis of digital images. By employing machine learning methods, we describe the current and emerging trends in image forensics, identifying the key issues regarding image tampering detection and attribution. In this paper, we establish a justification of the use of AI methods in the manipulation of forged images, with the aim of using such findings to help in cybercrime investigations. In addition, identifying the pros and cons of the integration of AI technologies within forensic analysis, we also highlight the ethical concerns and legal consequences that arise with the use of AI technologies in analysis. Overall, this work aims to progress the knowledge regarding AI utilization in the context of cybersecurity and offer suggestions for future study of this essential field.