test/add-tests-for-feature-engineering

[juniorcl]

Test: Test for Feature Engineering

Motivation

Enable feature selection based on permutation importance for sklearn-compatible models.

To ensure this, all feature engineering components must:

Be fully compatible com a API do scikit-learn

Produce deterministic and testable outputs

Preserve feature names and shapes correctly, enabling downstream model interpretability (e.g., permutation importance)

Scope

Implement unit tests for feature engineering wrapper classes to guarantee correct behavior, sklearn compatibility, and safe integration in modeling pipelines.

The tests should validate:

• Interface compliance (fit, transform, fit_transform)

• Correct handling of pandas DataFrames and NumPy arrays

• Preservation or correct transformation of feature names

• Consistent output shapes and data types

• Compatibility with sklearn tools (e.g., Pipeline, permutation_importance)

Acceptance Criteria

CategoryEncoderWrapper

• Correctly implements fit, transform, and fit_transform

• Returns output with a consistent shape compatible with sklearn

• Preserves or exposes feature names (get_feature_names_out)

• Works correctly inside an sklearn.Pipeline

• Does not raise errors when used with permutation_importance

FunctionTransformerWrapper

• Correctly implements fit, transform, and fit_transform

• Applies the transformation function consistently

• Preserves index alignment when using pandas DataFrames

• Returns output with predictable shape and dtype

• Works correctly inside an sklearn.Pipeline

• Does not raise errors when used with permutation_importance