Prediction of Mental Health based on Social Media User Behavior

With the integration of social media into daily life, the behavior data of using the platform, such as use duration and use habits, is an important method to judge a person’s mental state. In addition, long-term use of mobile phones and insufficient sleep time will also affect people’s mental health. The traditional mental health services need to be actively sought, and the coverage is insufficient. How to find psychological problems through users’ daily digital behavior. This paper collected the digital behavior data of 500 social media users, such as screen usage time, application switching times, sleep time, anxiety level, etc., and analyzed the relationship between these behavior characteristics and mental health status by constructing a decision tree and a random forest model. Among them, anxiety level is the most important mental health risk index, and the mental state of people who sleep for about 7 hours is better. In addition, screen usage time and application switching frequency can also be used to judge people’s mental health to a certain extent. The prediction accuracy of the random forest model was 91%, and the AUC value was 0.93, which was significantly better than the simple decision tree model. Using daily behavior data to understand their own status, it can provide a reference for individuals to judge their own status, set up health reminders on the platform, and improve community psychological services.