The following table records the Decision Tree experiments performed during the project.
| Experiment ID | Model | Hyperparameters | Preprocessing | Feature Selection | Train/Test Split | Precision | ROC-AUC |
|---|---|---|---|---|---|---|---|
| EXP-01 | Decision Tree | Default | None | All Features | 80/20 | 0.97 | 0.97 |
| EXP-02 | Decision Tree | Max Depth = 5 | Scaling | Selected Features | 80/20 | 0.98 | 0.98 |
| EXP-03 | Decision Tree | Max Depth = 7 | Scaling + Imputation | Selected Features | 80/20 | 0.99 | 0.99 |
| EXP-04 | Decision Tree | Pruned Tree | Scaling | Selected Features | 80/20 | 0.98 | 0.99 |
The first experiment established the baseline using a Decision Tree with default parameters and all available features.
Results
- Precision: 0.97
- ROC-AUC: 0.97
The second experiment introduced a maximum tree depth of 5 together with preprocessing and selected features.
Results
- Precision: 0.98
- ROC-AUC: 0.98
The third experiment used a maximum depth of 7 together with scaling, missing-value imputation, and selected features.
Results
- Precision: 0.99
- ROC-AUC: 0.99
This produced the highest recorded Precision among the experiments.
The fourth experiment evaluated a pruned Decision Tree configuration with preprocessing and selected features.
Results
- Precision: 0.98
- ROC-AUC: 0.99
The manually tracked experiments demonstrate how changes to model configuration, preprocessing, and feature selection can be recorded and compared systematically during machine learning development.