# Heart Disease Classification

## Docs

- [Model Architecture](https://heart.metesahankurt.cloud/docs/architecture.md): A 3-hidden-layer neural network for binary classification on tabular data.
- [Dataset](https://heart.metesahankurt.cloud/docs/dataset.md): UCI Heart Disease dataset — 1,025 patients, 13 clinical features, binary target.
- [Activation Functions](https://heart.metesahankurt.cloud/docs/experiments/activation-functions.md): Comparing Sigmoid, Tanh, ReLU, and Leaky ReLU in hidden layers.
- [Batch Size](https://heart.metesahankurt.cloud/docs/experiments/batch-size.md): How batch size affects update frequency, noise, and generalization.
- [Learning Rate](https://heart.metesahankurt.cloud/docs/experiments/learning-rate.md): Finding the right learning rate — from divergence at 0.1 to slow convergence at 0.0001.
- [Optimizers](https://heart.metesahankurt.cloud/docs/experiments/optimizers.md): Comparing Adam, SGD, RMSProp, and Adagrad convergence and accuracy.
- [Getting Started](https://heart.metesahankurt.cloud/docs/getting-started.md): Set up the environment and run the notebook locally.
- [Introduction](https://heart.metesahankurt.cloud/docs/introduction.md): Binary classification with deep learning on the UCI Heart Disease dataset.
- [Results](https://heart.metesahankurt.cloud/docs/results.md): Best configuration, overfitting analysis, and clinical interpretation of precision and recall.

## OpenAPI Specs

- [openapi](https://heart.metesahankurt.cloud/api-reference/openapi.json)