MindTime: Deep Learning Approach for Borderline Personality Disorder Detection

Borderline personality disorder (BPD) is a mental illness characterized by a long-term pattern of unstable relationships and strong emotional reactions. Those affected often engage in self-harm and other dangerous behavior. This paper introduces the MindTime system, which aims to detect Borderline personality disorder symptoms and signs of Self-harm ideation. MindTime is a personal diary mobile app. It allows users to make notes of daily events, experiences, thoughts, and feelings. The system features include password protection, mood tracking, text journals, voice recording (with speech to text), and video recording. Once the user adds a diary input, it is analyzed to detect if there are signs of BPD symptoms. We investigated different classifiers to extracts features from the stored diaries based on the input type text and video. The used classifiers were Naive Bayes, SVM, KNN, and finally LSTM. For text, The result showed that SVM and LSTM classifiers achieved the best user accuracy of 90.1% and 91%. As for audiovisual input CNN achieved user accuracy of 65%.