Development of intelligent system based on synthesis of affective signals and deep neural networks to foster mental health of the Indian virtual community

Social network analytics in health-care is recently catching-all as world is recovering from Covid-19 pandemic and when it comes to dealing with mounting mental health problems in the online community. A technical breakthrough to quickly address this issue is a big bet on researchers to deliver better intelligent mental health scale and support in mental distress risk management. In this paper we aim to build four artificial-intelligence driven models by blending the power of two dominant deep learning neural networks for explaining and predicting mental distress risk in the crowd-sourced online community. The models are simulated using a novel feature construct, which is a combination of numerical and textual data. The numerical data are realized by encoding social networking sites behavior and physical, social, cognitive experiences as part of three axes of psychological distress (depression, anxiety and stress). The textual part of data is made up of social network comments mined to acquire mental distressed cues by applying feature extraction techniques such as word embeddings and glove embeddings techniques. With the hyper-parameter tuning of models, we attained excellent performance (accuracy ~ 99.20%) which proves the efficacy of the proposed hybrid mental distress prediction model well with respect to accuracy in comparison to other related recent existing models with a boost of 0.20%. Our experimental results offer a robust model wherein we achieved zero false positive cases, attained 100% precision and excellent F1 measures. Additionally, we validated our results by using state-of-the-art technique BERT on different ground truth dataset i.e.,, Indian tweets to explore the tweets for psychological-distress prediction. Thus, we present an effectual automated tool for treatment activation of mental distress and supporting decisions in fostering the mental health of online society.