Identifying arrhythmias based on ECG classification using Enhanced-PSO method

Abstract

Electrical activity generated through the Electro-Cardio-Gram (ECG) was of vast quantities of data that are produced from the individual. The study of these data enables us through the classification of cardiovascular anomalies, to diagnose disease symptoms and disorders. The classification models are extremely dynamic and representative of the usage of computing tools in the unbalanced data environments is critical, where classes are not organized equally. Besides, the efficiency of classifying the ECG it was centered on evaluating the parameter and procedures for generating a very detailed, responsive, and reliable model. Only a few optimization parameters are considered in the existing model Convolution Neural Network (CNN). Instead of an analytical tuning of data level parameters, it chooses an algorithm level which leads to poor accuracy in the unbalanced data. Therefore, the machine needs careful attention, including metrics of accuracy and precision rate. This research work proposes a metaheuristic method Enhanced Particle Swarm Optimization (EPSO) for estimating parameters for unbalanced data arrhythmias classification. To define the arrhythmia form, here it chooses an unbalanced subset from the database. It should be processed with feature extraction and feature selection approach to combine the under sampling to solve the unbalanced data issue before classification.