Using decision tree to predict mental health status based on web behavior

It is very important to understand web users' psychological characteristics, which can help people adapt this rapid and complicated internet world better. Nowadays, web behavior plays a significant role in people's life and mental health. It is widely accepted in clinical psychology that mental activities and status are expressed in a way of behavior such as preference and choice. Web behavior, as a part of behavior, is supposed to be a meaning to detect users' mental status. In other words, individuals' mental status or moods are showed in these details of Internet usage behavior. In this paper, we propose to build a decision tree model to find out the relationship of web users' mental health and their behavior on the web. Our group conducts a questionnaire test to enroll subjects that use Internet frequently, and then extracts typical behavior features that could be mapped to virtual society and used to predict the users' mental health. Accordingly, subjects are also required to finish the Psychological Health Inventory (PHI) questionnaire as the criterion variables we are going to predict. Finally we get a collection of behavior features of importance to predict 7 mental disorders and the precision and recall are fairly good.