A Comprehensive Model to Monitor Mental Health based on Federated Learning and Deep Learning

Mental health is equally treated as important as physical health. Sound mental health leads to a peaceful life. Mental has a big impact on our thoughts, feelings, and behaviors. People’s mental health could be disturbed like facing depression. Depression is a major concern nowadays. People like to share their feelings and thoughts using several social media like Facebook, Twitter, WhatsApp, etc. In this paper, we propose a model based on federated learning and deep learning combined to monitor mental health using these social media data. In the proposed system data is collected from the user’s keyboard as people use the keyboard to type their thoughts, feelings on social media. Depression level is detected on daily basis using federated learning and recurrent neural network (RNN). The global model is saved into the global server. User’s local device inherits global model to test their daily used data on the keyboard. After testing, the user’s test data is sent anonymously to the global dictionary and then the global dictionary is updated daily using all user’s anonymous tested data. Then using this updated global sentiment dictionary global model is trained again and sent to all user’s local devices to monitor their mental health. Our proposed model acquires 93.46% accuracy on 60th day.