ReliefF based feature selection and Gradient Squirrel search Algorithm enabled Deep Maxout Network for detection of heart disease

Abstract

Detecting heart disease is challenging in clinical settings, leading to an increase in mortality rates. Current detection processes often rely on Electrocardiography (ECG) signal analysis, which requires accurate data processing and feature extraction. Traditional methods have limitations like processing time and accuracy. To address these issues, a novel approach called Gradient Squirrel Search Algorithm-Deep Maxout Network (GSSA-DMN) is proposed for more effective heart disease detection. The proposed GSSA-DMN approach involves several steps. Initially, input data is obtained from a specific database and subjected to data pre-processing, including log scaling for pattern transformation. Feature selection is then performed using ReliefF on the pre-processed data. The core of the approach lies in the Deep Maxout Network (DMN) trained by the Gradient Squirrel Search Algorithm (GSSA), which combines Gradient Descent Optimization (GDO) with the Squirrel Search Algorithm (SSA). The GSSA-DMN approach demonstrates remarkable performance. It achieves high accuracy, sensitivity, and specificity values of approximately 93.2%, 93%, and 91.5%, respectively. These results indicate its effectiveness in heart disease detection.Comparatively, the proposed GSSA-DMN method outperforms existing techniques. Its accuracy surpasses those of other methods by margins of 6.97%, 5.79%, 4.50%, 3.43%, and 1.93% when compared to BF-PSO, Bi-LSTM-CRF, XGBoost, RLNNC, and DMOA-SqueezeNet for K-value. This suggests that GSSA-DMN provides superior accuracy in detecting heart disease. In summary, the GSSA-DMN approach presents a promising solution for improving the accuracy and efficiency of heart disease detection compared to traditional methods and existing state-of-the-art techniques.