Depression is one of the most common mental health disorders, and a large number of depressed people commit suicide each year. Potential depression sufferers do not consult psychological doctors because they feel ashamed or are unaware of any depression, which may result in severe delay of diagnosis and treatment. In the meantime, evidence shows that social media data provides valuable clues about physical and mental health conditions. In this thesis, I argue that it is feasible to identify depression at an early stage by mining online social behaviours. My approach, which is innovative to the practice of depression detection, does not rely on the extraction of numerous or complicated features to achieve accurate depression detection. Instead, I propose a novel classifier, namely, Inverse Boosting Pruning Trees (IBPT), which demonstrates a strong classification ability on two publicly accessible Twitter depression detection datasets. Subsequently, I propose a novel deep neural topic model, FastText Distributional Prior DocNADE with Attention Mechanism (fdp-DocNADEa) that collects interpretable topics from Twitter users’ posting content and generates representative document features for depression classification. To comprehensively evaluate the clustering capability of the fdp-DocNADEa, I use three real text datasets and the fdp-DocNADEa still obtains competitive results against several state of the arts techniques. Finally, I combine the two newly proposed methods to form an entire framework for Twitter depressed users classification and the results manifest that my proposed framework is promising for identifying Twitter users with depression.