A feature selection system for improving ECG arrhythmia diagnosis using optimized MLP and GOA

Abstract The great majority of cardiac patients die less often when heart illnesses are detected early thanks to computer-aided diagnostic (CAD) equipment. It is a difficult undertaking to identify heart irregularities since low variations in ECG signals may be difficult for the eye to precisely distinguish. This research proposes the GOA-MLP classification model, an effective combination classification model utilizing Grasshopper Optimization Algorithm (GOA) and Multi-layer perceptron (MLP) for ECG arrhythmia diagnosis. In this method, the Neighbourhood Component feature selection method is utilized in conjunction with the Discrete Wavelet Transform and higher-order statistics to extract features. The proposed approach to categorizing the five classes of heartbeat categories has been contrasted with conventional neural networks and SVM-RBF kernel functions. The accuracy of our suggested system s classification of arrhythmia classes is great (99.66%). The simulation results demonstrate that the GOA-MLP approach has a higher classification accuracy than both the SVM-RBF and the neural network classifier.