In the US alone, over 10% of the population is currently diagnosed with diabetes. This is an extremely high percentage, and most individuals are unable to track their overall health because of how hard it is to diagnose diabetes early on.
What it does
To address this issue, we trained an ML model and built an interactive web application where users can answer specific questions about their health and get a diabetes prediction score from a scale of 0-2. 0 means that the detector does not detect any diabetes in the person whereas a score of 2 means that the person is likely to have diabetes.
How we built it
We built this application using HTML, CSS, and Javascript. We trained our model using scikit-learn in Python with a Diabetes dataset from Kaggle, and we connected our frontend to our backend, and deployed our ML model to the web using Flask.
Challenges we ran into
One challenge we ran into was when we were trying to deploy our ML model and had to connect ou frontend and backend together. This was a challenge as none of us have worked with Flask before, and we ran into errors that involved having to change how the backend retrieved the user input in order to run the model. Although this was challenging, with some research and teamwork we were able to successfully connect the two pieces together.
Accomplishments that we're proud of
We are proud of how we trained our ML classifier model with around a 85% accuracy. There are an abundance of factors that can lead to and worsen diabetes, and in our model we were successfully able to diagnose the disease with just 14 features as input.
What we learned
From this project, each of us brought our strengths to the table, and it was a great way for all of us to learn from each other. All of us especially learned about how user input can be sent to the backend for further processing and how to display this output back to the user.
What's next for D-Squared (Diabetes Detection)
Next up, we will be implementing some UI changes and adding more navigation screens to the application so that users can benefit from more personalized feedback. As of now, we have the bulk of the ML training done, and we hope to give users more detailed feedback about their health in the future.
Log in or sign up for Devpost to join the conversation.