Health_app

🏥 AI Powered multifunctional Healthcare Hub

An end-to-end AI ecosystem that combines multiple machine learning models into a unified healthcare platform. Users input symptoms or queries → system delivers diagnoses, precautions, medication alternatives, risk scores, and instant medical answers.

---

✨ Features

🔍 Disease Prediction & Medical Recommendation

• Predicts diseases from symptoms using RandomForest Classifier
• Provides: medical description, precautions, medication suggestions, diet recommendations

💊 AI-Powered Drug Recommendation

• Finds alternative medicines using NLP + Cosine Similarity
• Matches based on ingredients and properties

❤️ Heart Disease Risk Assessment

• LightGBM & EasyEnsemble classifiers
• Inputs: age, BMI, smoking, medical history
• Output: personalized risk score + recommendations

🤖 MediBot - AI Health Assistant

• Mistral-7B-Instruct LLM via Hugging Face
• RAG (Retrieval-Augmented Generation) with FAISS vector database
• Fact-based, fast, relevant medical answers

---

🛠️ Tech Stack

Component	Technologies
ML Models	RandomForest, LightGBM, EasyEnsemble, Scikit-learn
NLP	TF-IDF, Cosine Similarity, Transformers
LLM & RAG	Mistral-7B, Hugging Face, FAISS
Backend	Python, FastAPI, Pandas, NumPy
Deployment	Docker, CI/CD (optional)

---

📁 Project Structure

Health_app/
│   ├── models/                       # Trained ML models
│   │   ├── first_feature_model.pkl   
│   │   ├── second_feature_model.pkl   
│   │   └── third_feature_model.pkl   
│   ├── medi_bot/                      # LLM chatbot implementation
│   │   ├── connect_memory.py          # Base memory connections
│   │   └── connect_memory_llm.py      # LLM-specific memory handling
│   └── utils/                          # Helper functions
│       └── heart_disease.jpg           # Heart visualization asset
├── data/
│   ├── datasets/
│   │  
│   └── vector_store/
│       ├── deb.faiss                   # FAISS database binary
│       └── index.faiss                  # FAISS index file
├── requirements.txt
├── Dockerfile
└── README.md

🚀 Quick Start (Conda)

# Clone repository
git clone https://github.com/Addisu-Amare/Health_app.git
cd Health_app

# Create conda environment with Python 3.9
conda create -n health_app python=3.9 -y

# Activate environment
conda activate health_app

# Install dependencies
pip install -r requirements.txt
 then   run main.py

---

🔌 API Endpoints

Endpoint	Method	Input	Output
/predict/disease	POST	Symptoms list	Disease + precautions + diet
/recommend/drug	POST	Drug name	Alternative medicines
/assess/heart-risk	POST	Age, BMI, smoking, history	Risk score + recommendations
/chat/medical	POST	User query	AI medical response

---

🧠 Model Details

First Feature Model: Disease Prediction

• Algorithm: RandomForest Classifier
• Features: 132 symptoms (one-hot encoded)

Second Feature Model: Drug Recommendation

• Algorithm: NLP + Cosine Similarity
• Features: TF-IDF vectors of drug ingredients

Third Feature Model: Heart Risk Assessment

• Algorithm: LightGBM + EasyEnsemble
• Features: Age, BMI, smoking, BP, cholesterol, etc. )

---

🤖 MediBot Architecture

User Query
    ↓
[connect_memory.py]  ←  FAISS Vector Store (deb.faiss + index.faiss)
    ↓
[connect_memory_llm.py]  ←  Mistral-7B-Instruct (Hugging Face)
    ↓
RAG-enhanced response
    ↓
User

---

🌍 Why This Matters

This system is designed for healthcare accessibility — especially in regions with limited medical infrastructure. It's:

• ✅ Deployable — not just a demo
• ✅ Evidence-driven — RAG ensures factual responses
• ✅ Scalable — API-first architecture
• ✅ Explainable — every recommendation comes with context

---

🤝 Contributing

Contributions welcome! Open an issue or submit a PR.

---

📄 License

MIT © Addisu Amare

---

📬 Contact

Addisu Amare
GitHub | LinkedIn | Email

---

Making healthcare smarter, faster, and accessible — one query at a time.