Heart Disease Classification
This project builds and evaluates a binary classification neural network to predict the presence of heart disease in patients. Using the Heart Disease UCI dataset, it systematically explores different architectural choices — activation functions, optimizers, learning rates, and batch sizes — to understand their real-world impact on model performance.Academic context: Mid-term project for a Deep Learning course, guided by Doç. Dr. Öğr. Üyesi Abdullatif KABAN.
What This Project Covers
Dataset
1,025 patient records with 13 clinical features from the UCI Heart Disease dataset.
Architecture
A 3-hidden-layer neural network designed for tabular binary classification.
Experiments
Systematic comparison of activation functions, optimizers, learning rates, and batch sizes.
Results
Best configuration achieves ~97% test accuracy with strong precision and recall.
Tech Stack
| Tool | Role |
|---|---|
| Python 3.10+ | Language |
| TensorFlow / Keras | Model training |
| scikit-learn | Preprocessing |
| pandas / NumPy | Data handling |
| Matplotlib / Seaborn | Visualization |
| Jupyter Notebook | Development environment |