C45TreePruner

Author	Lily Dong
Creation date	06/01/2011
Firing policy	all
Package	org.seasr.meandre.components.prediction.decisiontree.c45

DESCRIPTION

This module prunes a decision tree built by the C4.5 Tree Builder. Detailed Description: This module prunes a decision tree using a reduced-error pruning technique. Error estimates for the leaves and subtrees are computed by classifying all the examples of the Example Table. Both subtree replacement and subtree raising are used. Subtree replacement will replace a node by one of its leaves if the induced error of the replacement is less than the sum of the errors for the leaves of the node. Subtree raising will lift a subtree if the error for the raised subtree is less than the original. The complexity of pruning the tree is O(n (log n)2).References: C4.5: Programs for Machine Learning by J. Ross QuinlanData Type Restrictions: The Unpruned Root must be a DecisionTreeNode built by the C4.5 Tree Builder.Data Handling: This module will attempt to classify the examples in the Example Table N times, where N is the number of nodes in the tree.Scalability: This module will classify the examples in the Example Table at least once for each node of the tree. This module will need enough memory to hold those predictions.

INPUTS

Name	Description	Example
treeNode	Read the root node of the unpruned decision tree. The root node is of type org.seasr.meandre.support.components.prediction.decisiontree.c45.DecisionTreeNode.
exampleTable	Read the training data that was used to build the decision tree. The training data is of type org.seasr.datatypes.datamining.table.ExampleTable.

OUTPUTS

Name	Description	Example
treeNode	Output a decision tree node which is the root of the pruned tree. The node is of type ncsa.d2k.modules.core.prediction.decisiontree.c45.DecisionTreeNode.

PROPERTIES

Name	Description	Default value
verbose	Control whether debugging information is output to the console	true