datarobot-model-explainability

What problem does it solve?

It helps you understand why a DataRobot model makes certain predictions by computing SHAP-based insights, XEMP prediction explanations, anomaly explanation artifacts, and common diagnostics like ROC/lift/confusion.

Core Features & Use Cases

• SHAP explainability (primary path): Generate full-row SHAP matrices, per-row top-feature previews, aggregated feature importance, and SHAP distributions, optionally filtered with Data Slices.
• XEMP prediction explanations (secondary path): Produce XEMP-based per-row explanations when SHAP is unavailable or when XEMP is specifically required, following required prerequisites like Feature Impact computation and initialization.
• Model diagnostics and anomaly explanations: Retrieve ROC, lift, and confusion insights and compute time-series anomaly assessment explanations using AnomalyAssessmentRecord.

Quick Start

Use the datarobot-model-explainability skill to compute SHAP values for all features and all rows for a given model by asking for a ShapMatrix with entity_id set to the model ID.