Decision Tree ROC Analysis

ROC curve comparison and hyperparameter tuning for Decision Tree classifiers on two OpenML datasets.

Overview

This project trains Decision Tree classifiers on two OpenML binary-classification datasets (IDs 4534 and 44), each with both the entropy and gini splitting criteria. For every configuration it generates an ROC curve via 10-fold cross-validated probability predictions, computes the AUC score, and runs a GridSearchCV over min_samples_leaf to find the best hyperparameter. Four ROC-curve plots are saved as PNG files.

Requirements

• Python 3.8+
• scikit-learn >= 1.0
• matplotlib >= 3.5
• numpy >= 1.21

Installation

pip install -r requirements.txt

Project Structure

decision-tree-roc-analysis/
├── plot_roc_curves.py      # ROC curve plotting and tuning script
├── requirements.txt        # Python dependencies
├── .gitignore
└── README.md

Usage

python plot_roc_curves.py

Steps performed:

1. Downloads OpenML datasets 4534 and 44.
2. Trains 4 Decision Tree configurations (2 datasets x 2 criteria: entropy, gini).
3. Computes 10-fold cross-validated ROC curves and AUC scores.
4. Saves each ROC plot as roc_curve_{dataset_id}_{criterion}.png.
5. Runs GridSearchCV on min_samples_leaf and prints the best parameters for each configuration.

Results

Four PNG plots are produced in the working directory:

File	Description
roc_curve_4534_entropy.png	Dataset 4534, entropy criterion
roc_curve_4534_gini.png	Dataset 4534, gini criterion
roc_curve_44_entropy.png	Dataset 44, entropy criterion
roc_curve_44_gini.png	Dataset 44, gini criterion

AUC values and best GridSearchCV parameters are printed to stdout.

Author

Biswajeet Sahoo

License

MIT License