Halcyon

Inspiration

We were inspired by the idea of a machine learning algorithm being able to recommend certain activities based on your own previous experience. As you continue to use the app, the better the app becomes at recommending actions for the user.

What it does

Asks the user to input their mood with respect to Energy, Motivation, Happiness, and Creativity, before uploading some of their health data as well as inputs to a firebase real-time database, where a python script is executed to read and format the apple health output and train 4 machine learning models to be able to accurately guess which health data has the most positive effect to the respective inputs from the user.

How we built it

We built it by having separate members create the application and work on getting the inputs to the database, while I worked on setting up google datalab to code the python script. We then worked in tandem to write and format the data for the machine learning models, before being able to read the exported data from the model and returning it to the user.

Challenges we ran into

Integrating the python script with the firebase database, as well as a large amount of data pre-processing for the ML models. There were also numerous troubles with swift, as none of us were entirely fluent with the swiftUI framework.

Accomplishments that we're proud of

Finishing a project in a domain where none of us were very comfortable.

What we learned

Not to give up, and the importance of stepping outside one's comfort zone.

What's next for Halcyon

More data to train and test models, as well as more data categories taken from the phone for further customization.

Built With

Submitted to

Hack the North 2020++

Created by

I worked on the IOS backend data itegration aspects of the project using Swift and Firebase. The first part was programming automated Apple Health data retrieval, which required configuring application permissions. Then, along with the user inputted data, the Apple Health data was sent to a Realtime Database on Firebase. The Database structure was designed for organization and easy data retrieval for both the mobile app and Google Cloud (where the ML Algorithm resides). The final step was configuring the data retrieval of the ML Algorithm's results back to the device, which was done with a simple database query.

Bishneet Singh
I worked on the front end design and UI. The app was entirely coded in XCode using SwiftUI. I also took part in integration of the cloud database and machine learning outputs.

Taewoo Ko
I was responsible for creating the python function in google cloud datalab that when executed, trains 4 seperate Machine Learning Algorithms in series to be able to recognize and predict which apple health input would most positively affect the user input values for energy, creativity, motivation, happiness. Upon start, the code reads and sorts all the read data by day and eliminates outliers before testing and training the gradient boosted regressor models. The models are then tested on the users most recent input, in which values are returned to the firebase real time database.

joshua dolgin
Kalany Bissonette