Nature Inspired Algorithm MABC for Clustering and Classification of ECG Heart Beats, Using Time and Frequency Domain Features

A nature inspired algorithm Artificial Bee Colony (ABC) has been used in many different application areas within last decade. Modified ABC algorithm for clustering and classification has been applied on different datasets in previous studies of the authors. In this paper an improvement on fitness function is realized and this improved algorithm (MABC) was applied on ECG heart beats. Electrocardiogram signals obtained from MIT-BIH dataset. Total 8 different heart beat types N, j, V, F, f, A, a and R are classified. In order to achieve better classification accuracy, frequency domain features are used in addition to time domain features. Feature selection is done by using divergence analysis. General classification accuracy and sensitivity results of MABC are compared with other methods, linear nearest mean classifier (NMC) and Kohonen's self organizing map (SOM) classifier. The highest accuracy 97.18% on analyzed dataset has been achieved by using the MABC algorithm as developed in this study.