weka.classifiers.trees.ht.HNode Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of weka-dev Show documentation
The Waikato Environment for Knowledge Analysis (WEKA), a machine learning workbench. This version represents the developer version, the "bleeding edge" of development, you could say. New functionality gets added to this version.
There is a newer version: 3.9.6
Show newest version
/*
 *   This program is free software: you can redistribute it and/or modify
 *   it under the terms of the GNU General Public License as published by
 *   the Free Software Foundation, either version 3 of the License, or
 *   (at your option) any later version.
 *
 *   This program is distributed in the hope that it will be useful,
 *   but WITHOUT ANY WARRANTY; without even the implied warranty of
 *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *   GNU General Public License for more details.
 *
 *   You should have received a copy of the GNU General Public License
 *   along with this program.  If not, see .
 */

/*
 *    HNode.java
 *    Copyright (C) 2013 University of Waikato, Hamilton, New Zealand
 *
 */

package weka.classifiers.trees.ht;

import java.io.Serializable;
import java.util.LinkedHashMap;
import java.util.Map;

import weka.core.Attribute;
import weka.core.Instance;
import weka.core.Utils;

/**
 * Abstract base class for nodes in a Hoeffding tree
 * 
 * @author Richard Kirkby ([email protected])
 * @author Mark Hall (mhall{[at]}pentaho{[dot]}com)
 * @revision $Revision: 9707 $
 */
public abstract class HNode implements Serializable {
  /**
   * For serialization
   */
  private static final long serialVersionUID = 197233928177240264L;

  /** Class distribution at this node */
  public Map m_classDistribution = new LinkedHashMap();

  /** Holds the leaf number (if this is a leaf) */
  protected int m_leafNum;

  /** Holds the node number (for graphing purposes) */
  protected int m_nodeNum;

  /**
   * Construct a new HNode
   */
  public HNode() {
  }

  /**
   * Construct a new HNode with the supplied class distribution
   * 
   * @param classDistrib
   */
  public HNode(Map classDistrib) {
    m_classDistribution = classDistrib;
  }

  /**
   * Returns true if this is a leaf
   * 
   * @return
   */
  public boolean isLeaf() {
    return true;
  }

  /**
   * The size of the class distribution
   * 
   * @return the number of entries in the class distribution
   */
  public int numEntriesInClassDistribution() {
    return m_classDistribution.size();
  }

  /**
   * Returns true if the class distribution is pure
   * 
   * @return true if the class distribution is pure
   */
  public boolean classDistributionIsPure() {
    int count = 0;
    for (Map.Entry el : m_classDistribution.entrySet()) {
      if (el.getValue().m_weight > 0) {
        count++;

        if (count > 1) {
          break;
        }
      }
    }

    return (count < 2);
  }

  /**
   * Update the class frequency distribution with the supplied instance
   * 
   * @param inst the instance to update with
   */
  public void updateDistribution(Instance inst) {
    if (inst.classIsMissing()) {
      return;
    }
    String classVal = inst.stringValue(inst.classAttribute());

    WeightMass m = m_classDistribution.get(classVal);
    if (m == null) {
      m = new WeightMass();
      m.m_weight = 1.0;

      m_classDistribution.put(classVal, m);
    }
    m.m_weight += inst.weight();
  }

  /**
   * Return a class probability distribution computed from the frequency counts
   * at this node
   * 
   * @param inst the instance to get a prediction for
   * @param classAtt the class attribute
   * @return a class probability distribution
   * @throws Exception if a problem occurs
   */
  public double[] getDistribution(Instance inst, Attribute classAtt)
      throws Exception {
    double[] dist = new double[classAtt.numValues()];

    for (int i = 0; i < classAtt.numValues(); i++) {
      WeightMass w = m_classDistribution.get(classAtt.value(i));
      if (w != null) {
        dist[i] = w.m_weight;
      } else {
        dist[i] = 1.0;
      }
    }

    Utils.normalize(dist);
    return dist;
  }

  public int installNodeNums(int nodeNum) {
    nodeNum++;
    m_nodeNum = nodeNum;

    return nodeNum;
  }

  protected int dumpTree(int depth, int leafCount, StringBuffer buff) {

    double max = -1;
    String classVal = "";
    for (Map.Entry e : m_classDistribution.entrySet()) {
      if (e.getValue().m_weight > max) {
        max = e.getValue().m_weight;
        classVal = e.getKey();
      }
    }
    buff.append(classVal + " (" + String.format("%-9.3f", max).trim() + ")");
    leafCount++;
    m_leafNum = leafCount;

    return leafCount;
  }

  protected void printLeafModels(StringBuffer buff) {
  }

  public void graphTree(StringBuffer text) {

    double max = -1;
    String classVal = "";
    for (Map.Entry e : m_classDistribution.entrySet()) {
      if (e.getValue().m_weight > max) {
        max = e.getValue().m_weight;
        classVal = e.getKey();
      }
    }

    text.append("N" + m_nodeNum + " [label=\"" + classVal + " ("
        + String.format("%-9.3f", max).trim() + ")\" shape=box style=filled]\n");
  }

  /**
   * Print a textual description of the tree
   * 
   * @param printLeaf true if leaf models (NB, NB adaptive) should be output
   * @return a textual description of the tree
   */
  public String toString(boolean printLeaf) {

    installNodeNums(0);

    StringBuffer buff = new StringBuffer();

    dumpTree(0, 0, buff);

    if (printLeaf) {
      buff.append("\n\n");
      printLeafModels(buff);
    }

    return buff.toString();
  }

  /**
   * Return the total weight of instances seen at this node
   * 
   * @return the total weight of instances seen at this node
   */
  public double totalWeight() {
    double tw = 0;

    for (Map.Entry e : m_classDistribution.entrySet()) {
      tw += e.getValue().m_weight;
    }

    return tw;
  }

  /**
   * Return the leaf that the supplied instance ends up at
   * 
   * @param inst the instance to find the leaf for
   * @param parent the parent node
   * @param parentBranch the parent branch
   * @return the leaf that the supplied instance ends up at
   */
  public LeafNode leafForInstance(Instance inst, SplitNode parent,
      String parentBranch) {
    return new LeafNode(this, parent, parentBranch);
  }

  /**
   * Update the node with the supplied instance
   * 
   * @param inst the instance to update with
   * @throws Exception if a problem occurs
   */
  public abstract void updateNode(Instance inst) throws Exception;
}