Boosting with feature selection technique for screening and predicting adolescents depression

Depression is a psychological disorder that is difficult to diagnose. Therefore, its early screening and accurate diagnosing are much needed. In this paper, we developed and evaluated a model that uses adolescents depression data for screening and predicting the depression based on boosting with feature selection techniques. These data have been collected from the Center for Epidemiologic Studies Depression Scale (CES-D) that is a short 20-item self-report scale. The model was trained and tested on a set of 3115 examples from Thasala Hospital. Our technique applied Max-Relevance Min-Redundancy (MRMR) feature selection to extract the principle feature items. MRMR technique can identify only twelve feature items that sensitive for predicting adolescents depression. Hence, our method uses only twelve sensitive feature items for training which differ from those prior works. In training step the data were classified by using the concept of AdaBoost algorithm based on decision tree classifier. Our results outperformed the other techniques under several performance evaluating functions. In addition, only twelve feature items are cover in four main elements of depression : emotional, cognitive, behavioral, and physical depression that particularly useful to screen and confirm to practitioners to ask questions and observe symptoms in adolescents depression clearly.