Emotion Detection of Mental Health Illnesses Using Physiological Data and Machine Learning Models

Abstract

Emotion Detection from physiological signals is a new field in Affective Computing in which Machine Learning (ML) models are vital. In this paper, the application of a hybrid model based on Random Forest and Support Vector Machine for emotion recognition is explored. Random Forest performs well in extracting intricate, non-linear relationships and handling high-dimensional data using ensemble learning, while SVM aims at optimizing the boundary decision by maximizing the margin among emotional classes. The hybrid classifier takes advantage of the strengths of both classifiers with enhanced prediction precision and generalization. In addition, the implementation of stacked ensemble learning improves the overall performance of the model. Experimental outcomes reveal that the suggested hybrid model (H-ML) obtains maximum accuracy of 94.32%, which is higher than separate Random Forest (91.105%) and SVM (91.24%) models. In addition, the efficacy of the model in recognizing negative emotions such as anger, sadness, fear, and disgust is illustrated. The H-ML model is found to perform better consistently, indicating its feasibility in identifying mental health problems arising due to sustained negative emotions. These findings support that suggested model provides significant advantage in emotion identification and has great potential for the diagnosis and treatment of mental diseases.