Al-Iraqia Journal for Scientific Engineering Research (Jun 2024)
Comparative Study of Different Machine Learning Algorithms to Classify EEG Commands
Abstract
Since electroencephalograms are non-invasive, inexpensive, and have high resolution, they are frequently used in diagnostic evaluations of a wide range of brain signals. However, manual EEG signal analysis can be exhausting and time-consuming. It takes a long time for physicians to become specialists in this field, and experts have low inter-rater agreement (IRA). To aid the last diagnosis and lessen the load, numerous Computer Aided Diagnostic (CAD) studies have looked into automating EEG interpretation. EEG signal classification provides an extra creative way of detecting emotions. Generic emotion recognition algorithms may face limitations and obstacles when restricting facial expression triggers and emotion masking. This study involves categorizing EEG data and evaluating the outcomes of several machine learning algorithms, such as Support Vector Machine (SVM), K-nearest Neighbor (KNN), Random Forest, Gradient Boosting, Logistic Regression, and Decision Trees. Grid search was also employed to reduce execution time for each of the machine learning models that were tested on the Spark cluster using hyperparameter tuning. This study used the Emotion Dataset, a multimodal dataset for classifying human affective states. Gradient Boosting outperformed other algorithms in terms of precision, recall, and F-Score, achieving 99.54%, 99.51%, and 99.52%, respectively, with an accuracy of 99.53%. According to the suggested model, several classification techniques are needed to differentiate between different emotional states, and gradient boosting is the most effective machine-learning technique. Current supervised classification techniques, including GBMs, are used to simulate a variety of typical machine-learning problems with high reliability. The recommended approach produced better accuracy and faster training speeds.
Keywords
