• clustering analysis
• LIWC
• mental health
• online peer support communities
• predictive analysis
• reddit
• user profile

Mental health issues are serious health problems still covered by social stigma and, for this reason, people dealing with these issues often do not feel free to share their conditions with those around them. One way they can overcome this problem is through online peer support communities where they can find people facing similar situations and ask for help or share their experiences without the fear of being judged. The aim of this thesis is to propose a psychometric user profile that can represent members of some of these communities on Reddit, based on the posts they have written, that can capture the differences between users dealing with addictions, anxiety, and depression. Besides, an unsupervised clustering approach is adopted to observe the data-driven subdivision of users according to their psychometric profile. Eventually, predictive models are built to study if the proposed profile can be used as an effective proxy to predict the category of mental health problems the user is struggling with