Doctor Diabetes

• 

  Form for user

• 

  Prediction positive

• 

  Prediction negative

• 

  Accuracy on test data

Inspiration

Using historic data, we can derive a lot of insights. My inspiration behind this project is to create a publicly available tool which can be used to generate diagnosis using real world, verified medical data.

What it does

Its simple! You input information and hit submit. We give you a result instantly!

How I built it

I built this application using web technologies and machine learning. I used a dataset i obtained from Kaggle

Frontend: The front end contains an HTML form which the user is expected to fill out. Once the user clicks on submit, the data flows to the backend.

Backend: The backend consists of a Flask Server.

• On initializing the server (i.e. start-up), a K-Nearest Neighbors classifier is trained using historic medical data.
• When the user inputs information in the front end, the data flows to the backend where a prediction is generated based on the pre-trained model.
• This prediction is then sent to the front end for the user to view

Challenges I ran into

• The initial idea was to create a web application to predict whether or not a user may have COVID-19. However I was unable to find a dataset that would work for this task.

What's next for Doctor Diabetes

This is a working model of a web application that is possible. To make this a reality there are several challenges:

• We will need reliable data. This means that the data needs to be labelled by medical professionals, anonymous and needs to be obtained from actual patients.
  
• We need a larger amount of data. The dataset I have used is a small dataset and there is no way to check if this information is reliable
• The model accuracy needs to be increased. I obtained an accuracy of 88% on unseen test data. To make this a reality, the accuracy needs to be higher.

Built With

• css
• flask
• html
• k-nn
• python
• scikit-learn