The Detection of State of Mental Health using Recurrent Neural Networks

Abstract

In the recent days, the number of people affected by Mental Depression Disorder (MDD) is on the rise with age, occupation related stress levels and several other factors. Depression has been identified as the main cause behind various diseases in individuals. In most cases, mental depression disorder is diagnosed with the help of counselling given by psychiatrists. However, even after the counselling and clinical diagnosis, the symptoms of depression persist. Social stigma associated with depression results in reluctance on the part of individuals to consult psychiatrists to diagnose mental illness. Also the existing techniques or methods do not guarantee accurate prediction of the level of depression. In order to overcome these problems, a new emotional model is designed to analyze the depression in individuals. A set of questionnaires called Personal Survey Questionnaire (PSQ) is framed to collect responses from the tweeters to understand about their mindset and depression level. Based on the PSQ answers, E-Ranking is calculated and compared with the polarity value generated by the PSQ answers. The performance of the proposed questionnaire-based model is compared with seven existing model based on parameters such as estimate and P-Value. Finally, the Recurrent Neural Network (RNN) is combined with Rule Based model (RB) to define the level and symptoms of depression. The blended RNN is compared with NLP process (Nature Language Processing) and it is proved that the Hybrid RNN and RB models give the best classification model for depression analysis.