HBP Risk Prediction – PU Learning

Overview

Project presents a Hypertension (HBP) Risk Prediction System using Positive-Unlabeled (PU) Learning. The system consists of:

• Machine Learning Model – Biased SVM trained to predict hypertension risk.
• Backend (Flask API) – Serves real-time predictions.
• Frontend (React.js + Tailwind CSS) – Web-based interface for user-friendly input.
• Deployment – Flask API on Render, Frontend on Cloudflare Pages.

🚀 Live Demo:

• Frontend: HBP Risk Prediction
• API Endpoint: HBP Risk API

PU Learning (Biased SVM Model)

• Instead of traditional supervised learning, this model learns from Positive (HBP cases) and Unlabeled (PU) data, where negative cases are uncertain. Weighted logistic pu learning as well as Pu bagging were trained but Biased svm showed robust results with prediction and was therefore chosen.
• Biased SVM assigns lower importance to unlabeled data to improve classification accuracy.

🚀 Deployment & API Usage

API Request

• The backend is deployed on Render with the API endpoint:
  https://pulearning-hbp-risk.onrender.com/predict
• The API accepts POST requests with JSON data in this format:

  {
      "features": [1, 52, 29, 0, 0, 0, 240, 115, 132, 84, 70, 77]
  }
  

API Response

 {
     "HBP Risk Probability": 0.81,
     "Risk Level": "High Risk (Hypertensive)"
 }