A Modified Deep Bi-Gated Recurrent Neural Network-Based Iot System for Effective Heart Disease Prediction

Majority of individuals are experiencing recurrent issues due to CAD, necessitating the development of an algorithm to accurately predict cardiac arrest onset. So, this paper proposes a heart disease prediction (HDP) model utilizing a deep learning modified neural network (EPHP-DLMNN). The solution makes use of the universal, accessible heart disease dataset, which contains patient-provided data on heart disease obtained via IoT sensors. In the preprocessing phase, adaptive median filter (SAMF) and fixed weighted mean (FWM) filtering are applied to the input image. Feature extraction is performed using methods like local binary pattern (LBP), gray level co-occurrence matrix (GLCM), gray level run length method (GLRLM), and Haralick features. The optimal features are selected using a hybrid African vulture with egret swarm optimization (HAVESO), combining African vulture optimization (AVO) and ESO. The selected features are classified by a modified deep bi-gated recurrent neural network (MDBi-GRNN), which is enhanced using a Bernoulli distribution function. The output is refined with the improved Beluga whale optimizer (IBWO) optimized by the butterfly optimization algorithm (BOA). The MDBi-GRNN model, implemented in Python, achieves superior performance with an accuracy of 0.984 and precision of 0.976 across various metrics including specificity, sensitivity, F-score, kappa value, and execution time.