Analysis of The Distance Metrics of KNN Classifier for EEG Signal in Stroke Patients

Abstract

The machine learning algorithms can be applied in many classification problems including the clinical studies. In the stroke emotion analysis, the machine learning is used to analyze the emotion of stroke patients and normal people. The K-Nearest Neighbor (KNN) classifier relies on the distance metric to calculate the nearest class for classification. The aim of this study is to compare the performances of different distance metrics apply on the classification of emotional electroencephalogram (EEG) between stroke and normal people. The Detrended Fluctuation Analysis (DFA) feature was extracted from the EEG signal of both classes and KNN was applied with different distance metrics for comparison. The results showed that the City Block distance metric performs the best among all. Moreover, the lower values of nearest neighbor were required to provide high classification result. The results from this study showed that the performance of the KNN classification was affected by the distance metric used.