Breath

Inspiration

As stress levels are rising in Singapore, even in young kids, there is a need to keep track of ones moods and physiological behaviours as they are key indicators of mental illnesses. We believe that there should be an app where such variables can be tracked and people are able to see if they are at risk mental illnesses such as depression. We strive to provide users a private space for their thoughts as well as a platform to receive advice to feel happier.

What it does

Breath is an application that accurately predicts a depressed individual and provides a safe space to deal with depression and stress healthily. It evaluates the initial risk a person has to depression based on demographic data and uses real-time information such as heart rate to compute the likelihood of stress.

How we built it

We created several machine learning models using both R and Python, finally settling for SGD Classifier and Random Forest models. These were used to predict the likelihoods of users suffering/being predisposed to depression and stress. Data collected to train the models ranged from the United States Centre for Disease and Control's databases as well as user data regarding physical indicators (such as heart rate and amount of sleep). This was the data used to train the various models, and we learned quite a lot.

The application was developed primarily in Flutter, and we created high fidelity prototypes beforehand in Figma so that everybody would be on the same page.

Challenges we ran into

The challenges that we ran into had to deal with mainly finding the data to supplement our building of the prediction models for both presence of depression based on demographic data as well as presence of stress which is based on sleep data and their corresponding heart rate.

Accomplishments that we're proud of

In just a day we were able to research and develop a model to predict if a person is at risk of depression based on variables such as recent issues with memory and problems sleeping at night.

We were also able to build a prototype on Figma and develop an early version using Android Studio.

What we learned

We learned, more than anything, the importance of time management. We were so focused on obtaining better data and fine tuning our model that our final submission took much longer than we expected.

Communication between team members was also very important, and we made full use of the Discord Platform - sharing screen and random interjections of ideas were very easy, and we learned to be more comfortable with one another. Vader's Fist will certainly live to fight another day.

It was certainly an experience to stay up and participate in this hackathon, and it has shown us that "much to learn, we still have".

What's next for Breath

Different people have a different thresholds for depression and stress. The model currently is a static model that predicts depression and stress. In the future, once we are able to get more information from the user we would be able to build a model that will learn from the patterns of the user and tailor the recommendations and solutions to individual users.