🩺 Heart Stroke Prediction Model

Overview

This repository contains a comprehensive data science project aimed at predicting the likelihood of a stroke based on user-provided clinical data. It integrates multiple tools and frameworks to build a full-fledged, production-ready application for healthcare professionals and medical devices.

Users

• Medical Professionals 👨‍⚕️
• Clinics & Hospitals 🏥
• Medical Devices 🔬

Usage Description 🫀

This application allows healthcare professionals or users to input clinical and personal health data through a web interface or directly into medical devices. Our machine learning model then predicts the probability of a stroke, displaying detailed results, precautions, and recommendations for professional consultation.

Features

• Web Interface & Data Search: Built with Streamlit for user-friendly interaction.
• Prediction API: Developed using FastAPI for real-time predictions.
• Machine Learning Model: Packaged as a Python module (stroke-pred-p0w11).
• Data Storage: Utilizes PostgreSQL with SQLAlchemy for secure and efficient data management.
• Data Ingestion: Implemented using Apache Airflow to collect and manage user input data.
• Monitoring Dashboard: Real-time prediction monitoring with Grafana.

Dataset

We utilize a dataset with 11 clinical features for predicting stroke events. You can access it here.

Getting Started

PostgreSQL Database Setup

1. Ensure you have the necessary dependencies installed:

   pip install -r Heart-Stroke-Prediction/requirements.txt

   Note: For Mac/Linux users, install psycopg2-binary instead of psycopg2.

2. Create a .env file in the root directory (Heart-Stroke-Prediction/.env) with the following content:

   [POSTGRES_DB]
   POSTGRES_USER=[Your_User]
   POSTGRES_PASSWORD=[Your_Password]
   POSTGRES_SERVER=[Your_Server]
   POSTGRES_PORT=[Your_Port]
   POSTGRES_DB=[Your_Database]
   
   [FastApi]
   BACKEND_SERVER=[Your_Server]

3. Navigate to the PostgreSQL setup directory and run the database setup script:

   cd stroke_heart_prediction/postgres
   python createdb.py

Features:

1. Data Ingestion with Airflow

2. Monitor with Grafana

3. Deploy the Web Interface on Heroku

Execute Locally

To run the application locally, execute the following commands:

# Start the Prediction API
cd Heart-Stroke-Prediction/stroke_api
uvicorn main:app --host 0.0.0.0 --port 8005

# Start the Web Interface
streamlit run web_interface.py --server.port 8010

System Architecture: