How does Scikit-Learn handle categorical encoding internally?

Scikit-Learn converts categorical data into numeric representations using encoders like OneHotEncoder or OrdinalEncoder.

Scikit-Learn does not natively process raw categorical strings in most estimators. Encoders map categories to numeric vectors. OneHotEncoder creates sparse binary matrices, while OrdinalEncoder assigns integer values. Proper encoding must be done inside pipelines to avoid leakage and ensure consistent category mapping.