Application of Graph Theory Features towards EEG Data Classification Models for Working Memory and The Emotional States

Abstract

Functional Connectivity analysis using Electroencephalography signals is a common 2 practice. The EEG signals are converted to networks by transforming the signals into a correlation 3 matrix and analyzing the resulting networks. Here, four learning models, namely, Logistic Regres 4 sion, Random Forest, Support Vector Machine, and Recurrent Neural Networks, are implemented 5 on the correlation matrix data to classify them either on their psychometric assessment or the 6 effect of therapy. The classifications based on RNN provided higher accuracy( 74-88%) compared 7 to the other three models( 50-78%). The use of a correlation matrix, instead of using individual 8 graph features provides an initial test of the data. When compared with the time-resolved correlation 9 matrix it provided 4-5% higher accuracy.