Studying Borderline Personality Disorder Using Machine Learning

Abstract

Borderline Personality Disorder is a mental disorder that impacts a person’s way of thinking and feeling about himself or others. This creates self-doubt, self-image problems, difficulty in managing emotions and behavior, as well as leads to unstable relationships. But this disorder can be cured with proper treatment if diagnosed early. The aim of this research is to explore potential features in the detection of Borderline Personality Disorder using machine learning methods. Furthermore, it identifies the potential features(here emotions) which are responsible in the detection of Borderline Personality Disorder. Data were collected in two ways one from self declared user in social media(i.e. who declared themselves as Borderline Personality Disorder patients) and second by inviting people for a Borderline Personality Disorder screening test Permission has been taken to fetch their social media data for research. This research achieved an accuracy of 81.03% using Random Forest classification algorithm. Emotional features were extracted from each data point in the dataset to identify the potential features for Borderline Personality Disorder classification. Features such as nervousness, shame, and pain have been derived that can in turn contribute to a similar accuracy to about 81.25%. This attempt is to explain the heuristic explanation, how differently an individual with Borderline Personality Disorder reacts from other individuals. This research paves the way for identifying Borderline Personality Disorder among individuals.