Stress Detection Using Smart-Watch By Automated Machine Learning Approach

Abstract

Continuous vulnerability to stress is injurious for psychological and Physical health, but to tackle and with stand the stress, we should first detect the stress.In this paper we introduce a fast and high accurate method for continuous stress detection using the data collected by a commercial wrist device. Applying machine learning models to real-life data is complex. But by using automated machine learning it is easy to apply for real-life datasets. So we have used AUTOGLUON an automated machine learning model for stress detection. In this paper we propose two models for stress detection. One model uses heart rate information and activities like walking, working, cycling, driving, playing soccer, climbing stairs and doing lunch for classifying stress. And another model uses smartwatch FITBIT daily summary data like Body related information, Heart rate information during physical activity, physical activity information and sleep information for classification of stress level.