IDMBD: Intelligent Diagnostic Modelling of Bipolar Disorder at its early onset

Abstract The proposed study presents an Intelligent Diagnostic Modelling of Bipolar Disorder (IDMBD), which represents an underlying framework operating on a Mobile Computing Device (MCD). The approach is meant to acquire multi-modal heterogeneous features in order to profile the participants more effectively with the highest granularity. IDMBD adopts a standard scale for forecasting the criticality state of Bipolar Disorder (BD). The feature-based profiling is carried out based on five discrete attributes collected by MCD, which, after undergoing a series of analytical processing, is subjected to a deep neural network-based learning model. An extensive evaluation is being carried out in multiple combinations of both machine learning and deep learning models to find that IDMBD exhibits optimal performance when integrated with Random Forest and Decision Tree with respect to multiple performance metrics.