package jsat.classifiers.trees;

import java.util.List;
import java.util.Random;
import java.util.Set;
import java.util.concurrent.ExecutorService;
import jsat.classifiers.DataPointPair;
import jsat.utils.ModifiableCountDownLatch;
import jsat.utils.random.RandomUtil;

/**
 * An extension of Decision Trees, it ignores the given set of features to use-
 * and selects a new random subset of features at each node for use. 
*
* The Random Decision Tree supports missing values in training and prediction. * * @author Edward Raff */ public class RandomDecisionTree extends DecisionTree { private static final long serialVersionUID = -809244056947507494L; private int numFeatures; public RandomDecisionTree() { this(1); } /** * Creates a new Random Decision Tree * @param numFeatures the number of random features to use */ public RandomDecisionTree(int numFeatures) { setRandomFeatureCount(numFeatures); } /** * Creates a new Random Decision Tree * @param numFeatures the number of random features to use * @param maxDepth the maximum depth of the tree to create * @param minSamples the minimum number of samples needed to continue branching * @param pruningMethod the method of pruning to use after construction * @param testProportion the proportion of the data set to put aside to use for pruning */ public RandomDecisionTree(int numFeatures, int maxDepth, int minSamples, TreePruner.PruningMethod pruningMethod, double testProportion) { super(maxDepth, minSamples, pruningMethod, testProportion); setRandomFeatureCount(numFeatures); } /** * Copy constructor * @param toCopy the object to copy */ public RandomDecisionTree(RandomDecisionTree toCopy) { super(toCopy); this.numFeatures = toCopy.numFeatures; } /** * Sets the number of random features to and use at each node of * the decision tree * @param numFeatures the number of random features */ public void setRandomFeatureCount(int numFeatures) { if(numFeatures < 1) throw new IllegalArgumentException("Number of features must be positive, not " + numFeatures); this.numFeatures = numFeatures; } /** * Returns the number of random features used at each node of the tree * @return the number of random features used at each node of the tree */ public int getRandomFeatureCount() { return numFeatures; } @Override protected Node makeNodeC(List> dataPoints, Set options, int depth, boolean parallel, ModifiableCountDownLatch mcdl) { if(dataPoints.isEmpty()) { mcdl.countDown(); return null; } final int featureCount = dataPoints.get(0).getDataPoint().numCategoricalValues()+dataPoints.get(0).getDataPoint().numNumericalValues(); fillWithRandomFeatures(options, featureCount); return super.makeNodeC(dataPoints, options, depth, parallel, mcdl); //To change body of generated methods, choose Tools | Templates. } @Override protected Node makeNodeR(List> dataPoints, Set options, int depth, boolean parallel, ModifiableCountDownLatch mcdl) { if(dataPoints.isEmpty()) { mcdl.countDown(); return null; } final int featureCount = dataPoints.get(0).getDataPoint().numCategoricalValues()+dataPoints.get(0).getDataPoint().numNumericalValues(); fillWithRandomFeatures(options, featureCount); return super.makeNodeR(dataPoints, options, depth, parallel, mcdl); //To change body of generated methods, choose Tools | Templates. } private void fillWithRandomFeatures(Set options, final int featureCount) { options.clear(); Random rand = RandomUtil.getRandom(); while(options.size() < numFeatures) { options.add(rand.nextInt(featureCount)); } } @Override public RandomDecisionTree clone() { return new RandomDecisionTree(this); } }

Ads help maintain this website.