Classification of Brain Activity Patterns from EEG Data Using WEKA

The paper is devoted to the classification of EEG data using WEKA library. The issue of identifying the eyes state was chosen as an illustration for the analysis. A brief overview of scientific research related to the topic of machine learning from the EEG data is given. The foundations of data classification using the WEKA library are described. We have investigated 34 classification algorithms that could be used to solve the problem of identifying the eyes state. The research was organized as follows. First, each classifier was tested on two datasets. Second, for the algorithms that showed the best results, additional research was carried out on the other nine data sets. The best results for the nine datasets were obtained using the IBk and RandomForest classifiers. The first dataset was taken from the UCI machine learning repository database. The rest of the datasets were generated from the EEG records included in the EEG Motor Movement / Imagery dataset. For the experiment, 30 seconds of recordings with open and closed eyes for 10 subjects were extracted from EDF files. The EEG data were filtered with a bandpass filter in the alpha band range. Next, the lines with outliers were deleted. The threshold for detecting outliers was defined as 3 standard deviations from the median.