Stress Detection

Home / portfolio / Stress Detection

Stress Detection

Using Wearable Stress Data and Machine Learning to Detect Stress

This project’s goal is to create a classifier that uses data from a smart fitness wearable to predict user’s stress states. This is an important problem as mental health is a growing dilemna and this tool can help people keep track of their stress levels over time and also be used to get alerts in real-time to stay calm in stressful situations. It can also help detect problems like long-term stress and alert designated people and avoid unhealthy long-term stress. For example, a parent may be able to track their children’s stress levels overall so they can provide them with extra care and comfort in times when they are not feeling good or in adverse social situations. This is an interesting problem because it can help detect if you are depressed, over-worked, or any other type of stress that one might induce. In the age of data and the internet of things, this smart wearable can be easily marketed to people who care not only about their physical but also their mental health. It can be integrated into wearables like the Apple Watch, FitBit, and so many more.

Feel free to check out the notebook in which I perform all the data gathering, pre-processing, feature extraction, and modeling steps which can be found on GitHub. You can also check out the powerpoint presentation for a high level overview or see the poster that I presentated at the Rutgers Annual Computer Science Conference in January 2019 below.


Poster Presentation at Rutgers Annual CS Conference - January 2019

Example Plots of Sensor Readings