ANALYZING EEG SIGNALS FOR STRESS DETECTION USING RANDOM FOREST ALGORITHM

Abstract

Detection of stress using EEG signals has gained much interest because of monitoring and early intervention. As for the contribution of this research, a reliable method for stress identification has been suggested, using a random forest model to categorize stress levels from EEG signals. Data were filtered using a bandpass filter, Independent Component Analysis, and more so using the Z-score to remove outliers and poor signals. Data that has been cleaned from noise and outliers will go through a feature extraction process using Power Spectral Density (PSD). The result of PSD is the power of each frequency of the EEG signal. The number of features used is 20. Random Forest was chosen due to its high accuracy and robustness in handling complex, high-dimensional data, which is common in EEG analysis. Thus, the model obtained an accuracy level of 0.8571, thereby approving the tool’s efficiency in distinguishing between different degrees of stress. The computational efficiency of the model, with a classification time of 0.2762 seconds, demonstrates its feasibility for practical applications. Based on these findings, it can be concluded that the Random Forest algorithm can be used to integrate wearable technology and for offering suggestions and timely interventions for better mental health.