Research on prediction model of college students’ mental health status based on machine learning algorithm

Abstract

Rapid and accurate assessment of college students’ mental health status is an important task in college mental health education and is also the basis for accurate intervention and personalized education services in college psychological work. The traditional evaluation methods have some problems, such as low real-time evaluation, poor evaluation effect of single-modal data, and social approval response bias. Social media big data provides a new way to predict college students’ mental health status in real time and accurately. This paper constructs a mental health prediction model based on web text data. BERT language framework is applied to psychological trait prediction, and BERT bidirectional training mode and Transformer coding module are used to mine more complete contextual semantic features and longer distance contextual dependencies, which solves the problem of enhancement vector representation of psychological trait semantic features. At the same time, considering that the difference in the internal structure of the classifier may lead to different classification effects, the fully connected layer of BERT model and the classical random forest algorithm are, respectively, used as two different classifiers in the downstream classification task to compare the model effects. The results show that BERT-B model has the greatest superiority, the prediction accuracy rate reaches 97.25%, the accuracy rate reaches 96.66%, the recall rate reaches 99.61%, the F1 value reaches 98.11%, and the AUC area reaches 99.44%, which can effectively predict the mental health status. The importance of this study is that it not only provides a fast, accurate and cost-effective mental health assessment tool for colleges and universities but also provides strong data support for personalized intervention and precise implementation of mental health services. The successful application of this innovative model marks an important step in the deep integration and value mining of social media big data in the field of mental health research and has great significance in promoting the modernization and scientific education of mental health.