A comprehensive approach to the analysis of electroencephalographic (EEG) signals obtained from human brain and artificially synthesized using machine learning methods is presented. The main focus is on data preprocessing, including signal normalization and filtering, as well as application of various feature extraction methods, in particular, Fast Fourier Transform and Mel-Frequency Cepstral Coefficients. A comparative analysis of classification accuracy using logistic regression, random forest, gradient boosting, and recurrent neural network LSTM is performed. Special attention is given to the effect of filtering parameters on classification accuracy. The results show that filtering and proper tuning of model parameters significantly improve the accuracy of EEG signal classification, ensuring separation of real and synthetic EEG pools. The results and discussion may serve as a basis for further research in the field of biomedical signal analysis and processing.