Investigations on multiple classifiers performance in detection of alcoholic patient state from EEG signals

Abstract

Alcohol-related disorders pose significant challenges in healthcare, newlinenecessitating accurate and efficient methods for detecting the alcoholic state newlinein patients. This thesis focuses on investigating the performance of multiple newlineclassifiers in detecting the alcoholic patient state from electroencephalogram newline(EEG) signals. newlineEEG (electroencephalography) is a technique used to record and newlinemeasure electrical activity in the brain. It involves placing electrodes on the newlinescalp, which detect the electrical signals generated by the firing of neurons in newlinethe brain. These signals are then amplified, filtered, and recorded for analysis. newlineThe research involves the collection of EEG data from both alcoholic and newlinenon-alcoholic subjects in a controlled environment. newlineThe research utilizes the Alcoholic EEG dataset from the UCI KDD newlinearchive database. The dataset comprises 64 channels, 256 Hz, and 1-second newlineepochs. It includes labels indicating whether the subjects are categorized as newlineAlcoholic or Control. The dataset consists of data from 244 subjects, with 122 newlinebeing normal and 122 being alcoholic patients. The data was collected using newlinethe 10/20 International montage, with each subject having 120 trials. newlineTo ensure the reliability and validity of the data, appropriate data newlinepreprocessing techniques are employed, including artifact removal and feature newlineextraction. These techniques aim to enhance the quality and relevance of the newlineEEG signals for subsequent analysis. newlineThe dimensionality reduction techniques like Hilbert transform, Ridge newlineregression, Chi Square probability density function, Fast Fourier transform newlineand fuzzy C means clustering. The characteristics were analysed the extracted newlinefeatures with histogram, scatter plot, norm plot, average statistical parameters, newlineand entropy measures. newline