Analysis of predisposition to addiction with machine learning techniques using eeg signals

Abstract

Diagnosis of alcohol dependence with Electroencephalography (EEG) signals is an important issue both personally and society. Today, many people are affected by this addiction. It has physiological effects, especially the brain, heart and immune system, as well as psychological effects. EEG signals are used effectively to observe these effects. In this study, genetic predisposition to alcoholism is diagnosed using EEG signals. Firstly, data analysis was performed on the EEG signal data obtained through the database. Recursive feature selection is used. For the classification, Multilayer Artificial Neural Networks (MLPNN), 1D-Convolutional Neural Networks (CNN), XGBoost Algorithm (XGBA), Random Forest Algorithm (RFA), K-Nearest Neighbor Algorithm (K-NN) are used. It has been studied in Pyhton environment. Accuracy, precision, sensitivity and F1 Score are used for classification performance criteria. Algorithms are compared according to the running time. In terms of classification success, MLPNN and CNN gives the best results. In terms of running time of algorithms, XGBA is the fastest running algorithm.