An Ensemble-Based Machine Learning Model for Emotion and Mental Health Detection

Abstract

Recent studies have highlighted several mental health problems in India, caused by factors such as lack of trained counsellors and a stigma associated with discussing mental health. These challenges have raised an increasing need for alternate methods that can be used to detect a person’s emotion and monitor their mental health. Existing research in this field explores several approaches ranging from studying body language to analysing micro-expressions to detect a person’s emotions. However, these solutions often rely on techniques that invade people’s privacy and thus face challenges with mass adoption. The goal is to build a solution that can detect people’s emotions, in a non-invasive manner. This research proposes a journaling web application wherein the users enter their daily reflections. The application extracts the user’s typing patterns (keystroke data) and primary phone usage data. It uses this data to train an ensemble machine learning model, which can then detect the user’s emotions. The proposed solution has various applications in today’s world. People can use it to keep track of their emotions and study their emotional health. Also, any individual family can use this application to detect early signs of anxiety or depression amongst the members.