• 

  Front End

HealthMatch

Goal

We ask you some basic health questions and use the CDC 2017 survey data to predict diseases that you could have/develop based on users with similar answers.

Twist

The app is Tinder styled - diseases have a bio and a description with a set of Do's and Dont's. You can swipe right to learn more, or swipe left to skip.

Inspiration

Tinder for diseases.

What it does

We ask you some basic health questions and use the CDC 2017 survey data to predict diseases that you could have/develop based on users with similar answers. The twist is that diseases have a bio and a description with a set of Do's and Dont's. You can swipe right to learn more, or swipe left to skip.

How we built it

We imported the CDC dataset using Pandas on Jupyter Notebook and wrote a fully Connected Neural Network to predict health issues for a new user, based on their answers to a few questions. The front-end is Angular - Tinder themed to make it quircky. We use a REST API on the Jupyter notebook to communicate.

Challenges we ran into

Importing the 1 GB dataset too up half the time. We haven't worked with R/Big Data/Parallel Programming so we had to improvise with Python.

Accomplishments that we're proud of

Managing to import the dataset and getting a prediction accuracy of 73%

What we learned

We learned a lot of new things - Jupyter, a bit of CUDA, a LOT of Tensorflow.

What's next for HealthMatch

Improving the model to get better accuracy and make the Disease Profiles dynamic.

Built With

• angular.js
• keras
• python
• tensorflow