Live demo: https://smart-healthcare-assistant-vszt.onrender.com
A web app that predicts heart disease risk from patient clinical data. Built with Python and Flask as a personal project during my MSc AI studies at BTU Cottbus.
During my first semester at BTU, I wanted to apply what I was learning in class to a real problem. Healthcare prediction felt like a good fit — the data is publicly available, the problem is meaningful, and it covers the full ML pipeline from raw data to a working interface.
You enter a patient's clinical details (age, blood pressure, cholesterol, etc.) and the app predicts whether they are at risk of heart disease. Results are shown instantly with a simple risk indicator (low / high).
There is also a separate page showing how the model performs — confusion matrix, feature importance, ROC curve, and dataset distribution.
The Kaggle version of this dataset has 1025 rows but only 302 are unique. I removed the duplicates before training to avoid inflated accuracy scores. Without deduplication the model scores 99%, which looks impressive but is misleading. After cleaning, accuracy is 84% which is honest and consistent with published research on this dataset.
| Metric | Score |
|---|---|
| Train / Test Split | 80% / 20% |
| Test Accuracy | 83.61% |
| 5-Fold Cross Validation | 82.78% ± 4.02% |
| Precision | 84% |
| Recall | 84% |
| F1-Score | 84% |
| ROC-AUC | 0.879 |
Evaluation performed on a held-out test set and validated using 5-fold cross-validation. Trained both Random Forest and XGBoost. Random Forest performed slightly better and was saved as the final model.
- Python, Flask
- Scikit-Learn, XGBoost
- Pandas, NumPy
- Matplotlib, Seaborn
- HTML, CSS
Smart-Healthcare-Assistant/
├── templates/
│ ├── index.html
│ └── charts.html
├── static/
│ ├── confusion_matrix.png
│ ├── distribution.png
│ ├── feature_importance.png
│ └── roc_curve.png
├── screenshots/
│ ├── home1.png
│ ├── home2.png
│ ├── charts1.png
│ └── charts2.png
├── app.py
├── model.py
├── charts.py
├── explore.py
├── heart.csv
├── requirements.txt
└── README.md
git clone https://github.com/MettuSurendraReddy/Smart-Healthcare-Assistant.git
cd Smart-Healthcare-Assistant
pip install -r requirements.txt
python model.py
python charts.py
python app.pyThen open http://127.0.0.1:5000 in your browser.
UCI Heart Disease dataset (Kaggle version)
302 unique records after removing duplicates
13 input features, 1 binary target (heart disease yes/no)
This project is for educational purposes only and should not be used for medical diagnosis.
- How duplicate data silently inflates model accuracy and why deduplication matters before training
- The difference between training accuracy and cross-validated accuracy
- How to build an end-to-end ML pipeline from raw data to a working web application
- Why 84% honest accuracy is better than 99% misleading accuracy
- How to use Flask to serve ML model predictions via a web interface
I am currently doing my MSc in Artificial Intelligence at Brandenburg University of Technology (BTU Cottbus) and looking for an internship in machine learning or data science.