Using biosensors and machine learning algorithms to analyse the influencing factors of study tours on students’ mental health

College students nowadays will inevitably deal with stress. Personal emotional and behavioural responses may be extremely strong when faced with stress. One of the most prevalent sources of stress for college students throughout the globe is mental health issues connected to stress. However, there is a lack of research focusing on the impact of specific activities, such as study tours, on students’ mental health and how these activities can be monitored using advanced technologies. As a result of its ability to analyze, classify, and alert college students’ psychological data with high quality, deep learning and machine learning have recently found widespread application in college students’ mental health education and management. Moreover, the integration of biomechanics and biosensor data offers new insights into understanding the physical and psychological impacts of study tours on mental health. This can potentially promote the development of colleges’ mental health education programs. Hence, this study proposes the Biosensor-based and Deep Neural Network-based College Student Mental Health Prediction Model (BDNN-CSMHPM) for detecting the mental stress of college students during study tours. Using biosensor data, including EEG and biomechanical metrics, this model employs the most effective BDNN to categorize the mental health condition as normal, negative, or positive. Consequently, BDNN is utilized to categorize the gathered emotional and biomechanical information, and based on the classification outcomes, the emotional condition of college students is determined. Considering that different features might stand in for different elements in the original data, it is necessary to extract several biosensors features to represent the information in the original EEG data accurately. Second, fusing various features is essential in the auto-learn model integration method. Third, the BDNN is fed the combined features, resulting in emotion classification. The numerical outcomes demonstrate that the BDNN-CSMHPM model enhances the student’s mental health prediction ratio of 98.9%, accuracy ratio of 96.4%, emotion recognition ratio of 95.3%, Pearson correlation coefficient rate of 97.2% and psychological monitoring ratio of 94.3% compared to other popular methods.