Evaluation of mental health education effectiveness based on IoT and graph neural networks

The provision of mental health education provides a basis for improved self-awareness, emotional regulation, and positive behavioral development in learners. However, current evaluations of mental health education use static models that do not adequately represent the complex, dynamic interactions between learners and their educational environments. Therefore, this study develops a new approach for evaluating the effectiveness of mental health education based on Dynamic Sled Dog Optimizer (DSDO)-driven Enhanced Graph Neural Network (EGNN) methodology, which optimizes and classifies the Physiological, Behavioral, and Learning Interaction features derived from the IMHED-320 dataset. The proposed architecture helps in understanding the mental status of students and devising educational programs in institutions. The preprocessing starts with normalization of values using Min-Max Normalization. Next, adaptive outlier detection is used to identify & remove all detectable anomalous values based on the distribution of each feature. Feature Extraction & Dimensionality Reduction are done using Principal Component Analysis (PCA). The performance evaluation of the proposed DSDO-EGNN architecture is done on a Python platform. The obtained experimental results achieve 0.978 accuracy & 0.982 AUC for predicting physiological features; 0.963 accuracy and 0.962 F1 Score for predicting behavioral features; and 0.982 accuracy with 0.989 AUC for predicting learning interactions. This indicates that the DSDO-EGNN architecture consistently performed better than traditional evaluation methods across all performance metrics. The research work provides a scalable, efficient, and robust architecture specifically designed to estimate the effectiveness of Mental Health Education via IoT Technologies used in Higher Education Institutions.