Classification and Statistical Analysis of Schizophrenic and Normal EEG Time Series

Abstract

In this study, discrimination of normal and schizophrenic EEG is aimed by using lineer features with different classifiers. Fort his purpose, 1 minutes of EEG records through 16 channels were recorded from 39 normal and 39 schizophrenia patients and minimum, maximum, mean, standard deviation and median feautes were extracted from these records. k-neighbors, Multi-layer perceptron, support vector machines and Random forest classifier were applied to feature vectors extracted from each channel. Highest classification accuracy is reached to 99.95% in proposed work. While MLP seems to be best classifier, channel C4 is observed most relevant to discriminate schizophrenic EEG from healthy control group. As a result of independent sample t-test and Mann-Whitney U Test for the purpose of statistical analysis, there is a distinct statistical significance for whole channels.When considering proposed work, obtained results are so promising and make contributions to literatüre view according to related works.