A Deep Learning Ensemble Framework for Mental Health Prediction

Abstract

Deep learning algorithms are used in mental health prediction to identify trends that may indicate potential mental health issues. This project aims to predict the likelihood of different mental health conditions such as bipolar disorder, schizophrenia, depression, anxiety disorder, and post traumatic stress disorder, based on a range of factors, including social behavior, lifestyle choices, medical history, and physiological data, to enable timely cure. Our dataset contains medical data of patients, focusing on their mental health and related symptoms. Our research involved training several DL models, such as MLP, CNN-ANN, LSTM-GRU to predict the mental health disorders. Each model was carefully trained and tested using cross- validation to ensure accuracy and reliability. The MLP model has achieved an accuracy of 92.41% , the hybrid CNN- ANN model has achieved an accuracy of 92.94% and the hybrid LSTM-GRU model reached the highest accuracy of 96.4%, highlighting its ability to predict mental health disorders.