jsat.classifiers.trees

Class MDI

java.lang.Object

jsat.classifiers.trees.MDI

All Implemented Interfaces:

java.io.Serializable, TreeFeatureImportanceInference

public class MDI
extends java.lang.Object
implements TreeFeatureImportanceInference

Determines the importance of features by measuring the decrease in impurity caused by each feature used, weighted by the amount of data seen by the node using the feature.
This method only works for classification datasets as it uses the ImpurityScore class, but may use any impurity measure supported.

For more info, see:

• Louppe, G., Wehenkel, L., Sutera, A., & Geurts, P. (2013). Understanding variable importances in forests of randomized trees. In C. j. c. Burges, L. Bottou, M. Welling, Z. Ghahramani, & K. q. Weinberger (Eds.), Advances in Neural Information Processing Systems 26 (pp. 431–439). Retrieved from here
• Breiman, L. (2002). Manual on setting up, using, and understanding random forests v3.1. Statistics Department University of California Berkeley, CA, USA.

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
MDI()
MDI(ImpurityScore.ImpurityMeasure im)

Method Summary

Modifier and Type	Method and Description
<Type extends DataSet> double[]	getImportanceStats(TreeLearner model, DataSet<Type> data)

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

MDI

public MDI(ImpurityScore.ImpurityMeasure im)

MDI

public MDI()

Method Detail

getImportanceStats

public <Type extends DataSet> double[] getImportanceStats(TreeLearner model,
                                                          DataSet<Type> data)

Specified by:

getImportanceStats in interface TreeFeatureImportanceInference

Parameters:

model - the tree model to infer feature importance from

data - the dataset to use for importance inference. Should be either a Classification or Regression dataset, depending on the type of the model.

Returns:

a double array with one entry for each feature. Numeric features start first, followed by categorical features. Larger values indicate higher importance, and all values must be non-negative. Otherwise, no constraints are placed on the output of this function.