Web mining and mental health
Mabruk Abusaa, Joachim Diederich, Ayyoub Ajmi · Intelligent Agents, Web Technologies and Internet Commerce · 2004
Diagnostic categories in psychiatry can be broad and heterogeneous, e.g. schizophrenia, which includes very different symptoms. In order to arrive at diagnostic categories that are practical and task-relevant, text mining (clustering) techniques are used to extract diagnostic information from psychiatric reports and other texts, in this case Internet newsgroups. Clustering techniques applied to newsgroup messages from sci.philosophy and sci.psychology.psychotherapy let to the identification of Attention Deficit Hyperactivity Disorder (ADHD) as a suitable diagnostic category. In further experiments, ADHD is then the target for a number of machine learning experiments that step-by-step eliminate ADHD-related words from the input documents, and therefore, increase the complexity of the classification task. The results show very good accuracy (92% for support vector machines) and precision (87%) and moderate to low recall. Elimination of semantically related words to ADHD results in a minimal reduction of performance.