Intelligent Patient Risk Assessment & Agentic Health Support System

An AI-powered healthcare analytics system that predicts patient health risk using machine learning and extends into an intelligent agent-based assistant for structured health recommendations.

---

Project Overview

Early detection of patient health risk is essential for preventive healthcare. Manual risk evaluation is time-consuming and often inconsistent.

This system automates risk assessment using traditional Machine Learning models and provides a scalable architecture for intelligent health guidance.

The project is divided into two milestones:

Milestone 1 – ML-Based Risk Assessment (No GenAI)

Predicts patient health risk using classical ML algorithms.

Milestone 2 – Agentic Health Assistant

Extends the system using an agent workflow to generate structured health reports and recommendations.

---

Problem Statement

Healthcare institutions require efficient systems to:

• Identify high-risk patients early
• Provide interpretable risk scores
• Assist in preventive healthcare planning
• Maintain consistency in evaluation

This project builds a structured ML-based risk prediction system and extends it into an intelligent decision-support assistant.

---

Key Features

Milestone 1 – Machine Learning Risk Prediction

• Data cleaning & preprocessing
• Feature engineering
• Multiple ML models:
  • Logistic Regression
  • Decision Tree
  • Random Forest
• Model comparison & evaluation
• Risk score generation (0–1 probability)
• Risk classification:
  • Low Risk
  • Medium Risk
  • High Risk
• Confusion matrix & performance metrics
• Feature importance visualization
• Interactive Streamlit interface
• Public deployment

---

Milestone 2 – Agentic AI Extension

• Uses predicted risk output
• Generates structured health summary
• Provides preventive recommendations
• Suggests follow-up actions
• Medical disclaimer inclusion
• Optional Retrieval-Augmented Generation (RAG)

---

System Architecture

Milestone 1 Workflow

User Input
↓
Data Preprocessing
↓
Feature Engineering
↓
ML Model
↓
Risk Score + Category
↓
UI Display

---

Milestone 2 Workflow

User Input
↓
ML Risk Prediction
↓
Agent Workflow
↓
Knowledge Retrieval
↓
Structured Health Report
↓
UI Display / Download

---

Project Structure

Installation & Setup

Follow these steps to set up the project on your local machine.

Prerequisites

• Python 3.11 - Ensure you have Python 3.11
• pip - Package installer for Python

Step 1: Clone the Repository

git clone https://github.com/yourusername/Agentic-Health-Support-System.git
cd Agentic-Health-Support-System

Step 2: Create a Virtual Environment

It is recommended to create a virtual environment to avoid dependency conflicts.

On macOS/Linux:

python3 -m venv venv
source venv/bin/activate

On Windows:

python3 -m venv venv
venv\Scripts\activate

Step 3: Install Dependencies

Install all required packages using the requirements.txt file:

pip install -r src/requirements.txt

Step 4: Run the Application

Start the Streamlit application:

streamlit run src/app.py

The application will open in your default web browser at http://localhost:8501.

Additional Configuration

Environment Variables

If needed, create a .env file in the project root with the following variables:

GROQ_API_KEY

Data Files

Ensure the following directories exist with required data files:

• data/ - Contains training and test datasets

Troubleshooting

• Module not found errors: Make sure you've activated the virtual environment and installed all dependencies
• Port already in use: Use a different port with streamlit run streamlit_app.py --server.port 8502

---