predictorImportance

Estimates of predictor importance for XGBoost model

Since R2026a

Syntax

imp = predictorImportance(mdl)

Description

imp = predictorImportance(mdl) computes estimates of predictor importance for mdl by summing the estimates over all weak learners in the XGBoost model. imp has one element for each input predictor in the data used to train the model. A high value indicates that the predictor is important for mdl, with 0 representing the smallest possible importance.

example

Examples

collapse all

Estimate Predictor Importance

Open Live Script

Estimate the predictor importance for all variables in the ionosphere data.

This example includes a pretrained XGBoost classification model trained using the ionosphere dataset. The model was trained in Python and saved as a json file using model.save_model('trainedXGBoostModel.json').

load ionosphere
modelfile = "trainedXGBoostModel.json";
Mdl = importModelFromXGBoost(modelfile)

Mdl = 
  CompactClassificationXGBoost
               ResponseName: 'Y'
                 ClassNames: [0 1]
             ScoreTransform: 'logit'
                 NumTrained: 30
    ImportedModelParameters: [1×1 struct]


  Properties, Methods

Estimate the predictor importance for all predictor variables.

imp = predictorImportance(Mdl);
figure
bar(imp)
title("Predictor Importance Estimates")
ylabel("Estimates")
xlabel("Predictors")

Figure contains an axes object. The axes object with title Predictor Importance Estimates, xlabel Predictors, ylabel Estimates contains an object of type bar.

The 27th predictor has the most importance and some predictors have zero importance on the classification.

Input Arguments

collapse all

`mdl` — XGBoost model
`CompactClassificationXGBoost` model object | `CompactRegressionXGBoost` model object

XGBoost model, specified as a CompactClassificationXGBoost or CompactRegressionXGBoost model object imported with importModelFromXGBoost.

More About

collapse all