Harnessing Fog Computing for Accurate Arrhythmia Event Prediction: A Deep Learning Application Placement Model

Abstract

Traditional cloud-based arrhythmia detection systems face challenges such as latency, bandwidth limitations, and data privacy concerns, which can hinder their effectiveness in emergency situations. Moreover, existing machine learning approaches often rely on manual feature extraction, which can lead to information loss or computational complexity. To address these issues, this work proposes a deep learning model that can run on fog nodes, enabling distributed computing resources for real-time ECG data analysis and immediate arrhythmia detection. The model is implemented using a Python-based feed-forward neural network architecture, optimized with Stochastic Gradient Descent (SGD), and utilizes the ReLU activation function and Sparse Categorical Crossentropy loss function. The methodology involves analyzing cardiac event patterns from ECG data, designing the deep learning model architecture, selecting appropriate optimization techniques, and evaluating the model's performance using metrics such as accuracy, precision, recall, and F1-score. The model is trained and tested on the publicly available MIT-BIH Arrhythmia Database, which contains ECG recordings from 100 patients labeled with five distinct arrhythmia event categories. The experimental results demonstrate the proposed model's exceptional performance, achieving a mean accuracy of 99.2%, precision of 99.0%, recall of 99.1%, and F1-score of 99.3% in arrhythmia classification. The model exhibits high reliability and accuracy in identifying different types of arrhythmic events, including Normal, Supraventricular ectopic beat, Ventricular ectopic beat, Fusion beat, and Unknown beat.