Inspiration

It is extremely difficult for people to realize they have diabetes before symptoms arise. We normally go to a doctor when there is discomfort and at that point, it is often too late. This website is very approachable and there are only 4 parameters one needs to input to get a fairly accurate risk assesment.

What it does

Our model takes 4 inputs from the user which are: BMI, Age, Blood Sugar, and Blood Pressure. Using these 4 inputs, our randomforest model will return the user's diabetic risk as a percentage.

How I built it

The model was written in python and then exported using Pickle. We then used Flask to connect our webpage to the model. Our webpage was coded in HTML and CSS with a little Javascript for animations.

Challenges I ran into

The biggest challenge of this entire project was definitely learning how to code a machine learning model. Neither of our team members are in the computer science field so we relied on internet resources to teach us about using the random forest classifier.

Another challenge that we ran into was learning how to publish the app using DigitalOcean.The most experience any of us had up until this weekend with hosting websites was purely through GitHub Pages which we found out did not support flask applications.

Accomplishments that I'm proud of

We're proud of the fact that we were able to publish a fully decorated website along with a fully functional machine learning model attached to it.

What I learned

We both learned a lot about not only using python/HTML/CSS but also diabetes as we had to do a lot of research for our "Learning Resources" Page.

What's next for Machine Learning for Diabetic Screening

If possible, we'd like to gain access to better and larger databases, such as those from a research group or a local hospital, so that our model can be trained for a higher accuracy.

Built With

Share this project:

Updates