importModelFromXGBoost

Import pretrained XGBoost model into MATLAB

Since R2026a

Syntax

Mdl = importModelFromXGBoost(modelfile)

Description

Mdl = importModelFromXGBoost(modelfile) imports a pretrained XGBoost model from a JSON file modelfile. The XGBoost model can be a binary classification, multiclass classification or a regression model.

Examples

Import XGBoost Classification Model

Import a pretrained XGBoost classification model trained using the ionosphere dataset. The pretrained model is provided with this example.

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

The model is imported as a CompactClassificationXGBoost model object.

Use the dot notation to view the imported model parameters.

Mdl.ImportedModelParameters

ans = struct with fields:
            BaseScore: 0.6374
            Objective: 'binary:logistic'
              Booster: 'Tree'
    NumBoostingRounds: 30
     HasParallelTrees: 1
             IsBinary: 1
           NumClasses: 2

The parameters indicate it is a binary classification model trained using the 'Tree' booster. Convert the response data to a boolean array to match the imported model.

Y = (Y=="g");

Use the loss function to evaluate the classification loss of the model.

loss(Mdl,X,Y)

ans = single

0.0476

Import XGBoost Regression Model

Import a pretrained XGBoost regression model trained using the carsmall dataset to predict the fuel economy (MPG) of a car. The pretrained model is provided with this example.

The model was trained in Python using Cylinders, Displacement, Horsepower, and Weight as predictors. Then, it was saved as a json file using model.save_model('trainedRegressionXGBoostModel.json').

load carsmall
modelfile = "trainedRegressionXGBoostModel.json";
Mdl = importModelFromXGBoost(modelfile)

Mdl = 
  CompactRegressionXGBoost
               ResponseName: 'Y'
          ResponseTransform: 'none'
                 NumTrained: 30
    ImportedModelParameters: [1×1 struct]


  Properties, Methods

The model is imported as a CompactRegressionXGBoost model object.

Use the dot notation to view the imported model parameters.

Mdl.ImportedModelParameters

ans = struct with fields:
            BaseScore: 23.7181
            Objective: 'reg:squarederror'
              Booster: 'Tree'
    NumBoostingRounds: 30
     HasParallelTrees: 1

The parameters indicate it is a regression model trained using the 'Tree' booster.

Use the trained model to predict the fuel economy (MPG) for a four-cylinder car with a 200-cubic inch displacement, 150 horsepower, and weighing 3000 lbs.

predict(Mdl,[4 200 150 3000])

ans = single

24.0842

Predict Multiclass Labels Using XGBoost Model

Import a pretrained XGBoost multiclass classification model trained using the NLP dataset with 13 classes. The pretrained model is provided with this example.

The model was trained in Python and saved as a json file using model.save_model('trainedMulticlassXGBoostModel.json').

load nlpdata
modelfile = "trainedMulticlassXGBoostModel.json";
Mdl = importModelFromXGBoost(modelfile)

Mdl = 
  CompactClassificationXGBoost
               ResponseName: 'Y'
                 ClassNames: [0 1 2 3 4 5 6 7 8 9 10 11 12]
             ScoreTransform: 'softmax'
                 NumTrained: 390
    ImportedModelParameters: [1×1 struct]


  Properties, Methods

The model is imported as a CompactClassificationXGBoost model object.

Use the dot notation to view the imported model parameters.

Mdl.ImportedModelParameters

ans = struct with fields:
            BaseScore: 0.5000
            Objective: 'multi:softmax'
              Booster: 'Tree'
    NumBoostingRounds: 30
     HasParallelTrees: 1
           NumClasses: 13
             IsBinary: 0

The parameters indicate it is a multiclass classification model trained using the 'Tree' booster. Convert the response data Y to an encoded array to match the imported model and the predictor matrix X to a full matrix.

labels = categories(Y);

% Create mapping with 0-based indexing
labelMap = containers.Map(labels,0:length(labels)-1);

% Encode Y
Y = cellstr(Y);
encodedY = zeros(size(Y));
for i = 1:length(Y)
    encodedY(i) = labelMap(Y{i});
end

fullX = full(X);

Use the predict function to evaluate the class for the ninth sample.

predict(Mdl,fullX(9,:))

ans = 
0

XGBoost Classification with Missing Predictor Values

Import a pretrained XGBoost classification model trained using the ionosphere dataset. The pretrained model is provided with this example.

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

The model is imported as a CompactClassificationXGBoost model object. Convert the response data to a boolean array to match the imported model.

Y = (Y=="g");

Create a test sample with 5 missing predictor values.

missingIdx = randi(size(X,2),[1,5])

missingIdx = 1×5

    28    31     5    32    22

for i=missingIdx
    X(1,i)= NaN;
end

Predict the class of the sample with missing predictor values.

predict(Mdl,X(1,:))

ans = 
1

Input Arguments

modelfile — Name of XGBoost model file

character vector | string scalar

Name of the XGBoost model file, specified as a character vector or string scalar. modelfile must be in the current folder, or you must include a full or relative path to the file. The XGBoost model must be pretrained and saved in the serialized JSON format. For more information on how to save a model, see Save XGBoost Model in Python.

Example: "xgboostModel.json"

Output Arguments

Mdl — Compact XGBoost model

CompactClassificationXGBoost model object | CompactRegressionXGBoost model object

Compact XGBoost model, returned as either a CompactClassificationXGBoost or CompactRegressionXGBoost object. The compact model does not include the data used to train the model. To reference properties of Mdl, use dot notation. For example, to access or display the cell vector of weak learner model objects for the underlying tree ensemble, enter Mdl.Trained at the command line.

Limitations

• The following features in XGBoost library are not supported for imported models :-

  • Multilabel classification

  • Categorical predictors

• importModelFromXGBoost is tested up to XGBoost version 3.1.

More About

Save XGBoost Model in Python

importModelFromXGBoost accepts XGBoost models in the JSON format. Use the following command in Python^® before import to save a trained XGBoost model :-

model.save_model("modelfileName.json")

XGBoost models can be trained and saved in other languages such as R, C, C++ etc. However, importModelFromXGBoost has been tested only for Python imports.

Supported Objectives

importModelFromXGBoost supports models trained with the following objectives :-

• reg:squarederror

• binary:logistic

• multi:softmax

• multi:softprob

reg:squarederror, binary:logistic and multi:softprob are the default objectives for regression, binary classification and multiclass classification respectively.

Missing Value Support

Missing values support is enabled in the imported XGBoost models. This aligns with the default behavior of XGBoost to accept predictors with missing values.

References

[1] XGBoost. “XGBoost.” Accessed August 14, 2025. https://xgboost.readthedocs.io/en/stable/.

[2] XGBoost. “Introduction to Model IO” Accessed August 14, 2025. https://xgboost.readthedocs.io/en/stable/tutorials/saving_model.html.

Version History

Introduced in R2026a