Genetic-Fuzzy Hybrid Approach for Arrhythmia Classification

Abstract

Cardiac decision support system has become an effective tool for monitoring, classification and prediction of heart diseases. The common purpose of many researches is to improve its performances. In this paper, Genetic-Fuzzy hybrid approach ensuring accuracy is proposed. It is planned to classify ECG signals into five cardiac types, including, the Normal class (N), Paced class (P), Left Bundle Branch Block class (LBBB), Right Bundle Branch Block class (RBBB) and Premature Ventricular Contraction class (PVC). The main purpose deals with the optimization of a fuzzy system. In fact, the Genetic Algorithm (GA) is applied mainly for tuning the membership and rules parameters. The Root Mean Square Error (RMSE) is considered as the cost function. Accordingly, the investigated approach presents efficient accuracy (RMSE = 0.398) when the Fuzzy Inference System (FIS) is of TakagiSugeno (TS) type and the Gaussian membership is selected. The good linguistic interpretation is the power of GeneticFuzzy hybrid approach regarding others in machine learning.