Diabetes Risk Prediction

Prediction of diabetes based on the signs and symptoms using machine learning algorithms.

Problem Definition

Diabetes mellitus, often known simply as diabetes, is a group of common endocrine diseases characterized by sustained high blood sugar levels. Diabetes mellitus is diagnosed with a test for the glucose content in the blood [wiki]. The dataset provides records of some signs and symthoms for people who have and have not diabetes.

This is a binary classification problem and our aim, in this study, is to come up with a supervised ML model which predicts diabetes for a person who provides the information with high accuracy.

Software And Tools Requirements

• Flask
• sklearn
• pandas
• numpy
• uvicorn

To install the required libraries, run the following command:

pip install -r requirements.txt

How to run it locally

Terminal

uvicorn main:asgi_app --port 10000

IDE

Create a python run configuration and choose the main.py file

Docker

1. Run docker build -t diabetes-prediction .
2. Run docker run -p 10000:10000 diabetes-prediction
3. Open your browser and go to http://localhost:10000/

Online

App is also available online: https://diabetes-risk-prediction.onrender.com/

Or as API: https://diabetes-risk-prediction.onrender.com/predict_api

{
    "Gender": "Female",
    "Polyuria": "No",
    "Polydipsia": "No",
    "sudden weight loss": "No",
    "weakness": "Yes",
    "Polyphagia": "No",
    "Genital thrush": "No",
    "visual blurring": "Yes",
    "Itching": "Yes",
    "Irritability": "No",
    "delayed healing": "Yes",
    "partial paresis": "No",
    "muscle stiffness": "No",
    "Alopecia": "Yes",
    "Obesity": "No",
    "Age": 48
}