Classification of Arrhythmia Beats Using Optimized K-Nearest Neighbor Classifier

Abstract

Artificial intelligence related technologies are outperforming present day screening methods in medical field. The classification of ECG beats to detect cardiac arrhythmia is of great significance in medical field. K- Nearest Neighbor (KNN) algorithm is a supervised instance-based learning algorithm. It is most popular non-parametric algorithm in data mining and statistics because of its simplicity and substantial classification performance. However, classification using KNN algorithm becomes complicated when the sample size and the feature attributes are large. This may reduce the performance of KNN classifier. For classifying arrhythmia beats, MIT-BIH arrhythmia database is considered. An Optimized K-Nearest Neighbor Classifier (O-KNN) is proposed in this paper and Simulation results are compared with the traditional KNN algorithm. A traditional KNN algorithm gives an accuracy of 96.98%. Optimizing hyper parameters, the accuracy of the optimized K-Nearest Neighbor (O-KNN) Classifier reaches 99.03%. The experimental results show that the proposed algorithm improves the classification accuracy of KNN classifier in processing large data sets.