Comparison of Machine Learning Optimization Techniques for EEG-Based Confusion Emotion Recognition

Machine Learning (ML) and Optimization techniques are considered two distinct fields of study in Computer Science. However, with the rapid development of Artificial Intelligence, these two fields are now frequently interacting with each other to enhance the performance of learning and searching/classifying capabilities of the algorithms. In this study, an optimization performance comparison of the current, widely used hyperparameter tuning algorithms on classifier models of confusion emotion using EEG signals. Two Evolutionary algorithms: Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) were compared with two Non-Evolutionary Algorithms: Random Search and Bayesian Optimization. The four algorithms were made to optimize a Support Vector Machine (SVM) which was trained to be a confusion emotion classifier using Wang et al's [19] dataset. Performance was assessed by calculating the prediction accuracy and error produced by the SVM that was trained using the yielded parameters by each hyperparameter tuning algorithm. Results showed that Evolutionary hyperparameter tuning algorithms were more effective than Non-Evolutionary algorithms because they yielded optimized parameters that produced the highest prediction accuracy and the lowest prediction error.