Integrative Approach for Precision Prediction of Chronic Kidney Disease: Anfis-Based Feature Selection and DCNN Classification

Abstract

The most dangerous disease in the world is chronic kidney disease (CKD). Identifying the disease is challenging when the doctor conducts various investigations without analyzing the facts. CKD data analysis is important for prediction and risk reduction. Furthermore, prior methodologies failed to pay attention to the mutual relations of the features and increased the dimensionality ratio, so the result produced inaccuracy. To resolve this problem, we propose an efficient approach using an ANFIS (Adaptive Network-Based Fuzzy Inference System) for feature selection and a Deep Convolution Neural Network Classifier (DCNN) for predicting CKD based on a deep neural network model. The Chronic Disease Impact Rate (CDIR) is estimated by taking into account the importance of the features affected by the medical margins identified. Using K-Cross Fold Validation (K_CFV), feature limit patterns are formed and validated to extract feature weight importance. The extracted features are selected with the support of the ANFIS to reduce the feature dimension. The selected features are trained with the Deep Convolution Neural Network Classifier (DCNNC) to classify chronic disease severity. The proposed model utilizes a large dataset of patient information to accurately identify individuals at high risk of CKD. The proposed approach has been demonstrated to be effective and efficient through experiments, outperforming traditional methods and achieving high prediction accuracy. Furthermore, the proposed model shows significant potential for early intervention and prevention of CKD.