Evaluation of EEG signal processing methods for performance optimization of predictive machine learning algorithms

Electroencephalogram (EEG) is a biomedical operation that detects brain activity via electrodes placed around the scalp, which measure the electric field generated by active brain cells. However, only a faint source of signals effectively reaches the EEG electrodes, imposing challenges to the development of predictive algorithms for time series forecasting of EEG signals. The present proposal seeks to investigate the effects of a variety of preprocessing and data analysis techniques across multiple EEG datasets, aiming to evaluate which methods better accommodate the data for training machine learning-based predictive time series algorithms, as well as inferring which methods and parameters may be generalized throughout the different datasets and what must be adapted to each situation.