agoel00 / HeartDiseasePredictionAPI

JSON API which returns predictions based on user input parameters using Scikit-learn and Flask.

Heart Disease Prediction API

Flask REST API which predicts probability of Coronary Heart Disease in a patient taking 9 parameters based on patient's history as input.

The API uses a Logistic Regression Model from scikit-learn trained on the Framingham Heart Study Dataset from Kaggle.

The model achieved a test accuracy of around 88%.

It is deployed on Heroku here.

View the Jupyter notebook here

/predict endpoint

• Takes 9 paramteres as input

• Returns a binary prediction (0 or 1) and probability as well.

  Sample query

    https://heartapi.herokuapp.com/predict?age=31&sex=1&cigs=5&chol=230&sBP=280&dia=0&dBP=90&gluc=87&hRate=84
  

  Sample output

  {
     "data":{
        "age":"31",
        "cigsPerDay":"5",
        "diaBP":"90",
        "diabetes":"0",
        "glucose":"87",
        "heartRate":"84",
        "sex":"1",
        "sysBP":"280",
        "totChol":"230"
     },
     "prediction":[
        1
     ],
     "probability":[
        [
           0.4587093009776524,
           0.5412906990223476
        ]
     ]
  }
  

/model endpoint

• Returns the model details such as intercept and coefficients.

    https://heartapi.herokuapp.com/model
  

Running locally

1. Clone the repository

       git clone https://github.com/agoel00/HeartDiseasePredictionAPI.git
    
      cd HeartDiseasePredictionAPI

2. Install dependencies

       pip install requirements.txt

3. Start the Flask server

       python3 app.py

A PWA which communicates with this API is deployed here

License

MIT