org.terrier.matching

Class JforestsModelMatching

java.lang.Object

org.terrier.matching.LearnedModelMatching

org.terrier.matching.JforestsModelMatching

All Implemented Interfaces:

Matching

public class JforestsModelMatching
extends LearnedModelMatching

Applies a Jforests regression tree learned model to a FeaturedResultSet. Learned model files are generated by Jforests. If you use this class, you are implicitly using the jforests library, and the applies.

Properties

• fat.matching.learned.jforest.model - filename of the ensemble model generated by jforests
• fat.matching.learned.jforest.statistics - filename of the feature statistics file generated by jforests

Since:

4.0

Author:

Craig Macdonald

Field Summary

Fields inherited from class org.terrier.matching.LearnedModelMatching

increment, logger, parent, score_is_feature

Constructor Summary

Constructors

Constructor and Description
JforestsModelMatching(Index _index, Matching _parent)
JforestsModelMatching(Index _index, Matching _parent, String modelFilename)
JforestsModelMatching(Index _index, Matching _parent, String modelFilename, Class<? extends edu.uci.jforests.learning.trees.Tree> treeClass)
JforestsModelMatching(Matching _parent)
JforestsModelMatching(Matching _parent, String modelFilename)

Method Summary

Methods

Modifier and Type	Method and Description
protected void	applyModel(int N, double[] in_scores, int featureCount, double[][] doubleFeatures, double[] out_scores) Apply the loaded learned model for identifying the top N documents.
protected void	loadModel(String model_filename)
protected void	loadModel(String model_filename, Class<? extends edu.uci.jforests.learning.trees.Tree> treeClass)
protected edu.uci.jforests.dataset.RankingDataset	makeDataset(int N, int featureCount, double[][] doubleFeatures)

Methods inherited from class org.terrier.matching.LearnedModelMatching

getInfo, match, setCollectionStatistics

Methods inherited from class java.lang.Object

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

Constructor Detail

JforestsModelMatching

public JforestsModelMatching(Index _index,
                     Matching _parent)
                      throws Exception

Throws:

Exception

JforestsModelMatching

public JforestsModelMatching(Matching _parent)
                      throws Exception

Throws:

Exception

JforestsModelMatching

public JforestsModelMatching(Matching _parent,
                     String modelFilename)
                      throws Exception

Throws:

Exception

JforestsModelMatching

public JforestsModelMatching(Index _index,
                     Matching _parent,
                     String modelFilename)
                      throws Exception

Throws:

Exception

JforestsModelMatching

public JforestsModelMatching(Index _index,
                     Matching _parent,
                     String modelFilename,
                     Class<? extends edu.uci.jforests.learning.trees.Tree> treeClass)
                      throws Exception

Throws:

Exception

Method Detail

loadModel

protected void loadModel(String model_filename)
                  throws Exception

Throws:

Exception

loadModel

protected void loadModel(String model_filename,
             Class<? extends edu.uci.jforests.learning.trees.Tree> treeClass)
                  throws Exception

Throws:

Exception

makeDataset

protected edu.uci.jforests.dataset.RankingDataset makeDataset(int N,
                                                  int featureCount,
                                                  double[][] doubleFeatures)

applyModel

protected void applyModel(int N,
              double[] in_scores,
              int featureCount,
              double[][] doubleFeatures,
              double[] out_scores)

Description copied from class: LearnedModelMatching

Apply the loaded learned model for identifying the top N documents.

Specified by:

applyModel in class LearnedModelMatching

Parameters:

N - - number of documents

in_scores - - sample scores

featureCount - number of features

doubleFeatures - indexed by feature THEN document

out_scores - - final scores