DATASET COMPUTE SHAP

Fits a surrogate model (currently linear) on the supplied dataset and computes SHAP (SHapley Additive exPlanations) feature-importance values for each input column. Use this worker to quantify how much each predictor contributes to the target response and to rank variable importance across a design space.

When to use

Classification: process.

Tagged: dataset, explainability, feature_importance, linear_model, sensitivity, shap, surrogate.

Inputs

Label ID Type Default Required Description
Dataset dataset_1 dataset   Input dataset (tabular) containing both predictor and response columns; must be a d3VIEW dataset object with at least the columns named in input_column and output_column.
Input Column input_column scalar   One or more column names (multi-select, drawn from dataset_1) that act as model features / independent variables for the SHAP analysis.
Output Column output_column scalar   Single column name (drawn from dataset_1) that represents the target / dependent variable the surrogate model is trained to predict.
Model type model_type scalar liearModel   Surrogate model family used before computing SHAP values; currently only ‘linearModel’ (polynomial linear regression) is supported — leave at default unless additional model types are added.
Model Params model_params textarea order=1   Key=value pairs (comma-separated, e.g. ‘order=1’) controlling model hyperparameters; ‘order’ sets the polynomial degree for the linear model — default ‘order=1’ gives a first-order linear fit.

Outputs

Label ID Type Description
SHAP values dataset_shap_1 dataset Tabular dataset of SHAP values, one row per observation and one column per input feature, quantifying each feature’s additive contribution to the model prediction.

Disciplines

  • ai_ml.prognosis
  • ai_ml.surrogate
  • data.statistics
  • design_exploration.sensitivity

Runnable example

A runnable example is registered for this worker. Open the example workflow on the d3VIEW canvas: /api/workflow/example?id=dataset_compute_shap


Auto-generated from transformation schema. Worker id: dataset_compute_shap. Schema hash: fb8974b938d8. Hand-curated docs in workerexamples/ override this page when present.