ce-data-preparation

Preprocess mixed data types by validating input shapes and automatically encoding categorical features to ensure reliable calibration within CE pipelines. This handles missing values and manages mappings for deterministic encoding across unseen categories.

Calibrated explanations pipelines fail on unseen categorical values without deterministic encoding. This skill validates input data and applies robust categorical handling, managing mappings to prevent errors when new categories appear during inference.

Yes, you can export and import preprocessing mappings for categorical encoding across different environments. This satisfies ADR-002 and ADR-009 requirements, allowing consistent mapping reuse and configurable preprocessing behavior throughout your workflows.

The best way to validate missing values and input shapes is to apply automated preprocessing that checks data types and structural integrity. This skill diagnoses input errors upfront, ensuring your dataset meets the robust validation requirements for CE pipelines.

Yes, this data preprocessing approach handles NaNs and categorical features simultaneously. It validates input data, automatically encodes categoricals, and manages missing values across datasets to resolve calibration and explanation challenges caused by mixed-type inputs.

No specific dependencies are required to validate and encode data for calibrated explanations workflows. This skill operates independently to validate input shapes, manage missing values, and encode categorical features without external library constraints.