An Efficient Parallel Ensemble Deep Classifier for the Prediction of Diabetic Disease

Early illness detection is crucial to enhancing the standard of healthcare and preventing serious health disorders before it is too late. In today's society, diabetes affects people of all ages and demographics. Diabetes increases the chances of developing renal illness, nerve damage, blood vessel damage, and blindness. It also contributes to heart disease. As a result, accurately assessing diabetes data is crucial. PEDC (Parallel Ensemble Deep Classifier) was created in this study for the efficient identification of diabetes illness with enhanced overall performances. In this work, there are five stages for predicting diabetic disease. In stage one; the pre-processing is done using the Max-Min normalization method. In stage two, IPSO-FCM (Improved Particle Swarm Optimization and FCM Algorithms) assesses medical data. The third method performs a hybrid feature extraction approach using AWBi-LSTM (Adaptive Weight Bi-Directional Long Short-Term Memory) to extract pertinent characteristics from various parts. Stage four proposes a ROA-based feature selection method (Remora optimization algorithm). In the end, a parallel ensemble-based classification model for diabetic illness prediction is suggested with better outcomes. To identify diabetic patients, three classifiers—Resnet V2, VGG 16 and MCNN (modified convolution neural network)—are introduced. In comparison to cutting-edge techniques, the simulation results show the suggested model to be the most accurate.