Main Content


Reduce size of regression tree model



ctree = compact(tree) returns a CompactRegressionTree version of the trained regression tree model tree. You can predict regressions using ctree exactly as you can using tree. However, because ctree does not contain training data, you cannot perform some actions, such as cross-validation.


collapse all

Compare the size of a full regression tree model to the compacted model.

Load the carsmall data set. Consider Acceleration, Displacement, Horsepower, and Weight as predictor variables.

load carsmall
X = [Acceleration Cylinders Displacement Horsepower Weight];

Grow a regression tree using the entire data set.

Mdl = fitrtree(X,MPG)
Mdl = 
             ResponseName: 'Y'
    CategoricalPredictors: []
        ResponseTransform: 'none'
          NumObservations: 94

Mdl is a RegressionTree model. It is a full model, that is, it stores information such as the predictor and response data fitrtree used in training. For a properties list of full regression tree models, see RegressionTree.

Create a compact version of the full regression tree—that is, one that contains enough information to make predictions only.

CMdl = compact(Mdl)
CMdl = 
             ResponseName: 'Y'
    CategoricalPredictors: []
        ResponseTransform: 'none'

CMdl is a CompactRegressionTree model. For a properties list of compact regression tree models, see CompactRegressionTree.

Inspect the amounts of memory that the full and compact regression trees consume.

mdlInfo = whos('Mdl');
cMdlInfo = whos('CMdl');
[mdlInfo.bytes cMdlInfo.bytes]
ans = 1×2

       12570        7067

ans = 0.5622

In this case, the compact regression tree model uses approximately half the memory that the full model uses.

Input Arguments

collapse all

Full regression tree model, specified as a RegressionTree model object trained with fitrtree.

Extended Capabilities

Version History

Introduced in R2011a