hospital-department-classifier

🏥 Hospital Agentic AI Assistant

Task-1: Agentic AI System | Task-2: ML-based Department Prediction

---

📌 Project Overview

This project implements a Hospital Agentic AI Assistant using a multi-agent architecture. The system supports:

• Patient login & profile management
• Natural language symptom queries
• Doctor availability lookup
• Appointment booking with SQL persistence
• ML-based department prediction from patient symptoms

The project is divided into two well-defined tasks:

• Task-1: Rule-based Agentic AI Assistant (core system)
• Task-2: Machine Learning model for symptom → department prediction

---

🧠 Task-1: Hospital Agentic AI Assistant

🎯 Objective

To design and implement a customer-facing agentic AI system that interacts with patients, manages doctor availability, and handles appointment booking using SQL.

---

🧩 Architecture (Multi-Agent System)

The system is built using specialized agents, each responsible for a specific task:

---

🗂 Database Design (SQLite)

doctors_availability

patients

---

🖥 Interfaces Supported

• Console-based interaction (app.py)
• Streamlit UI (streamlit_app.py)

---

▶️ How to Run Task-1

pip install -r requirements.txt
python app.py

OR (Streamlit UI)

streamlit run streamlit_app.py

---

✅ Example Interactions

• “I have chest pain”
• “Show available doctors in Cardiology on Monday”
• “Book a slot with Dr Rahul on Monday at 09:00-10:00”

---

🤖 Task-2: ML-Based Department Prediction

🎯 Objective

To build a machine learning model that predicts the relevant hospital department based on patient symptoms.

---

📊 Dataset

• Source: Kaggle Healthcare Dataset
• Total samples: 23,000+
• Final departments used (10 classes):

Cardiology
Dermatology
ENT
Endocrinology
Gastroenterology
General Medicine
Neurology
Orthopedics
Psychiatry
Pulmonology

---

🛠 Data Pipeline

• Raw dataset stored in: task2_ml/data/Healthcare.csv
• Cleaned dataset: symptom_department.csv
• Preprocessing logic: prepare_dataset.py

This ensures reproducibility and transparency.

---

🧠 Model Details

---

📈 Evaluation Metrics

Classification Report (summary):

• Accuracy: ~47%
• Macro F1-Score: ~0.30
• Weighted F1-Score: ~0.36

Confusion Matrix was generated to analyze misclassification trends and class imbalance.

Note: Some departments show lower precision due to overlapping symptom patterns and dataset imbalance, which is expected in real medical data.

---

▶️ How to Run Task-2 Locally

Inference

python task2_ml/inference.py

Evaluation

python task2_ml/evaluate.py

---

📁 Why Trained Model Is Not Uploaded?

The trained_model/ directory is intentionally excluded via .gitignore:

• Model size is large
• HuggingFace models are reproducible
• Training script is provided
• Industry-standard best practice

---

📂 Repository Structure

hospital-agentic-ai/
│
├── agents/          # Task-1 agents
├── auth/
├── database/
├── reports/
│
├── task2_ml/
│   ├── data/
│   ├── train.py
│   ├── inference.py
│   ├── evaluate.py
│   ├── model_info.txt
│
├── app.py
├── streamlit_app.py
├── schema.sql
├── requirements.txt
├── .gitignore
└── README.md

---

🚀 Key Highlights

• Clean multi-agent architecture
• SQL-based persistence
• Atomic booking transactions
• Scalable ML pipeline
• Clear task separation
• Industry-ready repo structure

---

🧾 Conclusion

This project demonstrates a complete end-to-end AI system combining:

• Rule-based agentic AI (Task-1)
• ML-based intelligence (Task-2)

The solution is modular, explainable, and extensible, making it suitable for real-world healthcare automation scenarios.