Abstract
This study aims to present a machine learning-based approach for detecting emotional states from Electroencephalogram (EEG) signals by utilizing multiple machine learning models with various parameter settings to achieve the best outcomes. Nine machine learning models, which are Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), k-Nearest Neighbors (kNN), Support Vector Machine (SVM), Light Gradient Boosting Machine (LGBM), Adaptive Boosting (AdaBoost), Multilayer Perceptron (MLP), and 1D Convolutional Neural Network (1D CNN), are employed in this study. A dataset consisting of EEG signals from 300 patients is employed to conduct the experiments. Additionally, multiple synthetic datasets of 20,000 data points are generated using Generative Adversarial Network (GAN), Synthetic Minority Over-sampling Technique (SMOTE), and Adaptive Synthetic Sampling (ADASYN). Both the real and synthetic datasets are utilized for training, testing, and validating the models. By comparing the performance of the models, it is determined that across 5 different datasets (Original, Original + GAN, Original + SMOTE, Original + ADASYN, Original + GAN + SMOTE + ADASYN), the MLP model achieves the highest accuracy and efficiency. They demonstrated a testing accuracy of 98.8% and a latency ranging from only 1.8ms - 4.8ms. The use of synthetic data in machine learning and deep learning models shortens the process and enhances accuracy. The results of this study are promising and hold potential benefits for physicians and healthcare professionals.