Heart disease prediction using ECG-based lightweight system in IoT based on meta-heuristic approach

Annually, the proportion of individuals suffering from cardiovascular disease rises significantly. Heart attacks are the most prevalent and unpleasant illness among them. Heart disease (HD) diagnosis can be complicated when there are multiple symptoms. The growing popularity of wearable smart devices has increased the likelihood of providing the Internet of Things (IoT). However, one of the biggest obstacles to overcome in implementing the system under IoT is developing a lightweight model for cardiac diagnosis and categorization. In this paper, we have presented a two-step heart disease classification method. This method includes demarcation of classes with the help of optimized non-linear support vector machine technique in the first step and determining the modified fuzzy class in the second step. Initially, pre-processing is accomplished using the ECG signals to eliminate noise and improve signal smoothness. Subsequently, features such as PQRS wave, linear characteristics, and reciprocal information are extracted from pre-processed signals. At the classification stage, the two-stage learning system is used to classify cardiac arrhythmias. First, using the wild horse optimization (WHO) technique (WHO-sigmoid-TH-NL-demarcation), each class is subjected to a binary classification based on feature demarcation, thresholding, and weighting of the sigmoid function. The information from the first stage will be transferred into the subsequent stage for an equal number of heart disease classifications. In the second step, a TS fuzzy logic system optimized by the Giza Pyramids Construction (GPC) approach (GPC-TS-Fuzzy) is utilized to classify each signal. The MIT-BIH arrhythmia dataset is used to assess the suggested approach. In a comprehensive evaluation of the suggested method, performance metrics including "accuracy, sensitivity, and specificity" yielded average values of 98.58 %, 98.13 %, and 96.47 %, respectively. The MATLAB platform is utilized to accomplish the proposed methodology.