Main Content


Transform new data using generated features



    NewTbl = transform(Transformer,Tbl) returns a table with transformed features generated by the FeatureTransformer object Transformer. The input Tbl must contain the required variables, whose data types must match those of the variables originally passed to gencfeatures or genrfeatures when Transformer was created.


    NewTbl = transform(Transformer,Tbl,Index) returns a subset of the transformed features, where Index indicates the features to return.


    collapse all

    Generate features to train a linear regression model. Compute the cross-validation mean squared error (MSE) of the model by using the crossval function.

    Load the patients data set, and create a table containing the predictor data.

    load patients
    Tbl = table(Age,Diastolic,Gender,Height,SelfAssessedHealthStatus, ...

    Create a random partition for 5-fold cross-validation.

    rng("default") % For reproducibility of the partition
    cvp = cvpartition(size(Tbl,1),KFold=5);

    Compute the cross-validation MSE for a linear regression model trained on the original features in Tbl and the Systolic response variable.

    CVMdl = fitrlinear(Tbl,Systolic,CVPartition=cvp);
    cvloss = kfoldLoss(CVMdl)
    cvloss = 45.2594

    Create the custom function myloss (shown at the end of this example). This function generates 20 features from the training data, and then applies the same training set transformations to the test data. The function then fits a linear regression model to the training data and computes the test set MSE.

    Note: If you use the live script file for this example, the myloss function is already included at the end of the file. Otherwise, you need to create this function at the end of your .m file or add it as a file on the MATLAB® path.

    Compute the cross-validation MSE for a linear model trained on features generated from the predictors in Tbl.

    newcvloss = mean(crossval(@myloss,Tbl,Systolic,Partition=cvp))
    newcvloss = 26.9429
    function testloss = myloss(TrainTbl,trainY,TestTbl,testY)
    [Transformer,NewTrainTbl] = genrfeatures(TrainTbl,trainY,20);
    NewTestTbl = transform(Transformer,TestTbl);
    Mdl = fitrlinear(NewTrainTbl,trainY);
    testloss = loss(Mdl,NewTestTbl,testY);

    Train a linear classifier using only the numeric generated features returned by gencfeatures.

    Load the patients data set. Create a table from a subset of the variables.

    load patients
    Tbl = table(Age,Diastolic,Height,SelfAssessedHealthStatus, ...

    Partition the data into training and test sets. Use approximately 70% of the observations as training data, and 30% of the observations as test data. Partition the data using cvpartition.

    c = cvpartition(Tbl.Gender,Holdout=0.30);
    TrainTbl = Tbl(training(c),:);
    TestTbl = Tbl(test(c),:);

    Use the training data to generate 25 new features. Specify the minimum redundancy maximum relevance (MRMR) feature selection method for selecting new features.

    Transformer = gencfeatures(TrainTbl,"Gender",25, ...
    Transformer = 
      FeatureTransformer with properties:
                         Type: 'classification'
                TargetLearner: 'linear'
        NumEngineeredFeatures: 23
          NumOriginalFeatures: 2
             TotalNumFeatures: 25

    Inspect the generated features.

    Info = describe(Transformer)
    Info=25×4 table
                                          Type        IsOriginal         InputVariables                                              Transformations                                      
                                       ___________    __________    ________________________    __________________________________________________________________________________________
        zsc(Weight)                    Numeric          true        Weight                      "Standardization with z-score (mean = 153.1571, std = 26.8229)"                           
        eb5(Weight)                    Categorical      false       Weight                      "Equal-width binning (number of bins = 5)"                                                
        c(SelfAssessedHealthStatus)    Categorical      true        SelfAssessedHealthStatus    "Variable of type categorical converted from a cell data type"                            
        zsc(sqrt(Systolic))            Numeric          false       Systolic                    "sqrt( ) -> Standardization with z-score (mean = 11.086, std = 0.29694)"                  
        zsc(sin(Systolic))             Numeric          false       Systolic                    "sin( ) -> Standardization with z-score (mean = -0.1303, std = 0.72575)"                  
        zsc(Systolic./Weight)          Numeric          false       Systolic, Weight            "Systolic ./ Weight -> Standardization with z-score (mean = 0.82662, std = 0.14555)"      
        zsc(Age+Weight)                Numeric          false       Age, Weight                 "Age + Weight -> Standardization with z-score (mean = 191.1143, std = 28.6976)"           
        zsc(Age./Weight)               Numeric          false       Age, Weight                 "Age ./ Weight -> Standardization with z-score (mean = 0.25424, std = 0.062486)"          
        zsc(Diastolic.*Weight)         Numeric          false       Diastolic, Weight           "Diastolic .* Weight -> Standardization with z-score (mean = 12864.6857, std = 2731.1613)"
        q6(Height)                     Categorical      false       Height                      "Equiprobable binning (number of bins = 6)"                                               
        zsc(Systolic+Weight)           Numeric          false       Systolic, Weight            "Systolic + Weight -> Standardization with z-score (mean = 276.1429, std = 28.7111)"      
        zsc(Diastolic-Weight)          Numeric          false       Diastolic, Weight           "Diastolic - Weight -> Standardization with z-score (mean = -69.4286, std = 26.2411)"     
        zsc(Age-Weight)                Numeric          false       Age, Weight                 "Age - Weight -> Standardization with z-score (mean = -115.2, std = 27.0113)"             
        zsc(Height./Weight)            Numeric          false       Height, Weight              "Height ./ Weight -> Standardization with z-score (mean = 0.44797, std = 0.067992)"       
        zsc(Height.*Weight)            Numeric          false       Height, Weight              "Height .* Weight -> Standardization with z-score (mean = 10291.0714, std = 2111.9071)"   
        zsc(Diastolic+Weight)          Numeric          false       Diastolic, Weight           "Diastolic + Weight -> Standardization with z-score (mean = 236.8857, std = 29.2439)"     

    Transform the training and test sets, but retain only the numeric predictors.

    numericIdx = (Info.Type == "Numeric");
    NewTrainTbl = transform(Transformer,TrainTbl,numericIdx);
    NewTestTbl = transform(Transformer,TestTbl,numericIdx);

    Train a linear model using the transformed training data. Visualize the accuracy of the model's test set predictions by using a confusion matrix.

    Mdl = fitclinear(NewTrainTbl,TrainTbl.Gender);
    testLabels = predict(Mdl,NewTestTbl);

    Figure contains an object of type ConfusionMatrixChart.

    Input Arguments

    collapse all

    Feature transformer, specified as a FeatureTransformer object.

    Features to transform, specified as a table. The rows must correspond to observations, and the columns must correspond to the predictors used to generate the transformed features stored in Transformer. You can enter describe(Transformer).InputVariables to see the list of features that Tbl must contain.

    Data Types: table

    Features to return, specified as a numeric or logical vector indicating the position of the features, or a string array or cell array of character vectors indicating the names of the features.

    Example: 1:12

    Data Types: single | double | logical | string | cell

    Output Arguments

    collapse all

    Transformed features, returned as a table. Each row corresponds to an observation, and each column corresponds to a generated feature.

    Version History

    Introduced in R2021a