Feature Selection Using Selective Opposition Based Artificial Rabbits Optimization for Arrhythmia Classification on Internet of Medical Things Environment

An Electrocardiogram (ECG) is a non-invasive test that is broadly utilized for monitoring and diagnosing the cardiac arrhythmia. An irregularity of the heartbeat is generally defined as arrhythmia, which potentially causes the fatal difficulties that creates an instantaneous life risk. Therefore, the arrhythmia classification is a challenging task because of the overfitting issue caused by high dimensional feature space of ECG signal. In this research, the incorporation of the Internet of Medical Things (IoMT) is developed with artificial intelligence to provide the health monitoring for people who are having arrhythmia. In this work, the time, time-frequency, entropy, nonlinearity features of ECG and deep features of ECG from Convolutional Neural Network (CNN) are extracted to obtain different categories of ECG signal features. The Selective Opposition (SO) strategy based Artificial Rabbits Optimization (SOARO) is proposed for selecting the optimal feature subset from the overall features to avoid the overfitting issue. The chosen features are used to improve the classification done by Auto Encoder (AE). Further, the Shapley additive explanations (SHAP) based model is used to interpret the classified output from AE. The MIT-BIH arrhythmia database is used for evaluating the proposed SOARO-AE. The performance of the proposed SOARO-AE is evaluated by using the accuracy, sensitivity, specificity, recall and F1-Measure. The existing researches such as C-LSTM, DL-LAC-CNN, CNN-DNN, MC-ECG, FC and MEAHA-CNN are used to evaluate the SOARO-AE method. The accuracy of SOARO-AE is 98.89% which is high when compared to the C-LSTM, DL-LAC-CNN, CNN-DNN, FC and MEAHA-CNN.