Abstract
This study addresses the mental health challenges faced by students majoring in sports education and explores more effective strategies for mental health education. Using Internet of Things (IoT) data mining, relevant datasets are collected and categorized. A Random Forest (RF) model is then trained and optimized through a genetic algorithm, resulting in the Genetic Algorithm-Random Forest (GA-RF) psychological state perception model. The model is evaluated against multiple classification approaches. In the depression dichotomy experiment, the GA-RF model achieves superior results, with an optimized accuracy and an F1 score of 0.81, outperforming other algorithms in psychological state perception. By applying this model, routine data from students' daily activities can be analyzed to provide timely insights into their mental health. These insights support adjustments to teaching content and offer schools an evidence-based approach to improving instruction. Overall, the GA-RF model enhances data mining and prediction of students' psychological states, making it a valuable tool for advancing mental health education in sports education programs.