Spaces:
Sleeping
Sleeping
| """ | |
| This module sets up a Decision Tree Classifier for hyperparameter tuning. | |
| Features: | |
| - Uses `DecisionTreeClassifier` from scikit-learn. | |
| - Defines a parameter grid suitable for both binary and multi-class classification. | |
| - Default scoring: 'accuracy'. | |
| Considerations: | |
| - `criterion`, `max_depth`, `min_samples_split`, and `min_samples_leaf` are common parameters to tune. | |
| - Ordinal encoding will be used for tree-based models if implemented, but the pipeline code decides that. | |
| """ | |
| from sklearn.tree import DecisionTreeClassifier | |
| estimator = DecisionTreeClassifier(random_state=42) | |
| param_grid = { | |
| 'model__criterion': ['gini', 'entropy'], | |
| 'model__max_depth': [None, 5, 10], | |
| 'model__min_samples_split': [2, 5], | |
| 'model__min_samples_leaf': [1, 2], | |
| # Preprocessing params | |
| #'preprocessor__num__imputer__strategy': ['mean', 'median'], | |
| #'preprocessor__num__scaler__with_mean': [True, False], | |
| #'preprocessor__num__scaler__with_std': [True, False], | |
| } | |
| default_scoring = 'accuracy' | |