🫀 Heart-Disease-Risk-Prediction

The Final Project of the GTC Machine Learning Internship that predicts heart disease risk using the UCI dataset. The pipeline covers preprocessing, modeling (supervised & unsupervised), evaluation, and a Streamlit web app with an ngrok tunnel for easy sharing.

---

▶ Presentation Link ◀

https://www.canva.com/design/DAG0AqCMFGE/YDVwuiIJjpIkyoXfbCeoGg/edit?utm_content=DAG0AqCMFGE&utm_campaign=designshare&utm_medium=link2&utm_source=sharebutton

📹 Demo Video:

https://drive.google.com/file/d/1bdSyFYLScatsLacx6i5Czd3NhzIuNESv/view?usp=sharing

---

🔍 Project phases

Short summary of what we did in the project:

• Preprocessing & Cleaning & EDA — missing values, type casts, distributions and initial visual exploration.
• Feature selection — tested multiple methods: Chi-square, RFE (recursive feature elimination), and Feature Importance from a Random Forest.
• PCA analysis — dimensionality reduction and visual inspection of principal components.
• Supervised learning — trained classifiers to predict target (heart disease), evaluated with probability scores and metrics.
• Unsupervised learning — clustering / exploratory methods to inspect structure in the data.
• Hyperparameter tuning — grid/random search to obtain the final tuned pipeline.
• Web Deployment — Deploying the final pipeline on streamlit + ngrok.

Final model artifact: final_tuned_pipeline.pkl (used by the Streamlit app).

---

🚀 Quick install & run (Streamlit + ngrok)

Replace values where needed (ngrok token, filenames). The examples below assume the repo name Heart-Disease-Risk-Prediction and the deploy script deploy_with_pyngrok.py.

1. Clone the repo

git clone https://github.com/Ali-Islam111/Heart-Disease-Risk-Prediction.git
cd Heart-Disease-Risk-Prediction

2. Create & activate a virtual environment

python -m venv .venv
.\.venv\Scripts\Activate.bat

3. Install requirements

pip install -r requirements.txt

4. Set your ngrok token (one of these options):

Export as env var (recommended):

$env:NGROK_AUTHTOKEN="YOUR_NGROK_AUTHTOKEN"

Or open the deploy script and set NGROK_AUTHTOKEN = "YOUR_NGROK_AUTHTOKEN" (not ideal for public repos).

5. Run using the deploy script (this starts Streamlit and opens ngrok):

python deploy_with_pyngrok.py

⚠️ Important notes

Ngrok tunnels are just proxies — the public URL works only while the process (Streamlit + ngrok) is running somewhere. If you stop the process or shut the machine down, the link becomes unreachable.

To keep the app online 24/7, deploy to a server/VM and run the script as a background service, or use Streamlit Community Cloud / Render / Railway for a hosted solution.

Free ngrok URLs are ephemeral and change on each run; paid plans allow reserved subdomains.

📂 Useful files

app.py — Streamlit frontend.

deploy_with_pyngrok.py — helper to start Streamlit + open ngrok tunnel (one-command launch).

final_tuned_pipeline.pkl — trained model pipeline used by the app.

02_heart_disease_preprocessed.csv — preprocessed dataset used for EDA & PCA.

---

Developed by:

• Ali Islam Taha
• Ahmed Salah
• Mohamed Sameh
• Hassan Assar
• Mahmoud Khattab
• Mohamed Esam