Inspiration

Diabetes is a major public health problem that is approaching epidemic proportions globally. People delay the test, by creating various kinds of excuses. I lost someone very close just because of this neglect. This app is made to give people an idea of how serious their situation is so that they ultimately get a doctor's opinion.

What it does

The web app takes 10 inputs and predicts the chance of having diabetes. It then suggests some remedies for certain input parameters. The input parameters are : age,polyuria,polydipsia,gender,partial_paresis,weight_loss,irritability,healing,alopecia,itching.

How we built it

The model is built using Random Forest Classifier. The web app is built using streamlit.

Challenges we ran into

  • Getting a good dataset with enough values for difficulty.
  • Trained the model on different algorithms but Random Forest Classifier ultimately gave an accuracy of 98.72%. I also referred to a research paper that is added to the Github repo

Accomplishments that we're proud of

  • The model is predicting well even on foreign data
  • The web app is smooth

What we learned

  • How to pickle a model.
  • How to find a correlation between features and the target.

What's next for Diabetes Predictor

Creating a real-time app that keeps track of the health of a person along with many other features.

Built With

Share this project:

Updates