bifreader.java

MacroWeka扩展了著名数据挖掘工具weka

/*
 * 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 2 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, write to the Free Software
 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

/*
 * BIFReader.java
 * Copyright (C) 2003 Remco Bouckaert
 * 
 */

package weka.classifiers.bayes.net;

import java.io.*;
import java.util.*;

import javax.xml.parsers.*;
import org.w3c.dom.*;

import weka.classifiers.bayes.BayesNet;
import weka.classifiers.bayes.net.estimate.DiscreteEstimatorBayes;
import weka.core.*;
import weka.estimators.*;

/**
 * Builds a description of a Bayes Net classifier stored in XML BIF 0.3 format.
 * See http://www-2.cs.cmu.edu/~fgcozman/Research/InterchangeFormat/
 * for details on XML BIF.
 * 
 * @author Remco Bouckaert ([e-mail])
 * @version $Revision: 1.1 $
 */


public class BIFReader extends BayesNet {
    private int [] m_nPositionX;
    private int [] m_nPositionY;
    private int [] m_order;

	/** processFile reads a BIFXML file and initializes a Bayes Net
	 * @param sFile: name of the file to parse
	 */
	public BIFReader processFile(String sFile) throws Exception {
		m_sFile = sFile;
        DocumentBuilderFactory factory = DocumentBuilderFactory.newInstance();
        factory.setValidating(true);
        Document doc = factory.newDocumentBuilder().parse(new File(sFile));
        doc.normalize();

        buildInstances(doc, sFile);
        buildStructure(doc);
        return this;
	} // processFile

	String m_sFile;
	public String getFileName() {return m_sFile;}


	/** buildStructure parses the BIF document in the DOM tree contained
	 * in the doc parameter and specifies the the network structure and 
	 * probability tables.
	 * It assumes that buildInstances has been called before
	 * @param doc: DOM document containing BIF document in DOM tree
	 */
    void buildStructure(Document doc)  throws Exception {
        // Get the name of the network
		// initialize conditional distribution tables
		m_Distributions = new Estimator[m_Instances.numAttributes()][];
        for (int iNode = 0; iNode < m_Instances.numAttributes(); iNode++) {
        	// find definition that goes with this node
        	String sName = m_Instances.attribute(iNode).name();
			Element definition = getDefinition(doc, sName);
/*
	        if (nodelist.getLength() == 0) {
	        	throw new Exception("No definition found for node " + sName);
	        }
	        if (nodelist.getLength() > 1) {
	        	System.err.println("More than one definition found for node " + sName + ". Using first definition.");
	        }
	        Element definition = (Element) nodelist.item(0);
*/	        
	        
	        // get the parents for this node
	        // resolve structure
	        FastVector nodelist = getParentNodes(definition);
	        for (int iParent = 0; iParent < nodelist.size(); iParent++) {
	        	Node parentName = ((Node) nodelist.elementAt(iParent)).getFirstChild();
	        	String sParentName = ((CharacterData) (parentName)).getData();
	        	int nParent = getNode(sParentName);
	        	m_ParentSets[iNode].addParent(nParent, m_Instances);
	        }
	        // resolve conditional probability table
		        int nCardinality = m_ParentSets[iNode].getCardinalityOfParents();
	        int nValues = m_Instances.attribute(iNode).numValues();
	        m_Distributions[iNode] = new Estimator[nCardinality];
			for (int i = 0; i < nCardinality; i++) {
				m_Distributions[iNode][i] = new DiscreteEstimatorBayes(nValues, 0.0f);
			}

/*
	        StringBuffer sTable = new StringBuffer();
	        for (int iText = 0; iText < nodelist.getLength(); iText++) {
	        	sTable.append(((CharacterData) (nodelist.item(iText))).getData());
	        	sTable.append(' ');
	        }
	        StringTokenizer st = new StringTokenizer(sTable.toString());
*/
	        String sTable = getTable(definition);
			StringTokenizer st = new StringTokenizer(sTable.toString());
	        
	        
			for (int i = 0; i < nCardinality; i++) {
				DiscreteEstimatorBayes d = (DiscreteEstimatorBayes) m_Distributions[iNode][i];
				for (int iValue = 0; iValue < nValues; iValue++) {
					String sWeight = st.nextToken();
					d.addValue(iValue, new Double(sWeight).doubleValue());
				}
			}
         }
    } // buildStructure

    /** synchronizes the node ordering of this Bayes network with
     * those in the other network (if possible).
     * @param other: Bayes network to synchronize with
     * @throws Exception if nr of attributes differs or not all of the variables have the same name.
     */
    public void Sync(BayesNet other) throws Exception {
    	int nAtts = m_Instances.numAttributes();
    	if (nAtts != other.m_Instances.numAttributes()) {
    		throw new Exception ("Cannot synchronize networks: different number of attributes.");
    	}
        m_order = new int[nAtts];
        for (int iNode = 0; iNode < nAtts; iNode++) {
        	String sName = other.getNodeName(iNode);
        	m_order[getNode(sName)] = iNode;
        }
    } // Sync

	/** getNode finds the index of the node with name sNodeName
	 * and throws an exception if no such node can be found.
	 * @param sNodeName: name of the node to get the index from
	 * @return index of the node with name sNodeName
	 */
    public int getNode(String sNodeName) throws Exception {
    	int iNode = 0;
    	while (iNode < m_Instances.numAttributes()) {
    		if (m_Instances.attribute(iNode).name().equals(sNodeName)) {
    			return iNode;
    		}
	    	iNode++; 
    	}
    	throw new Exception("Could not find node [[" + sNodeName + "]]");
    } // getNode

    /**
     * Returns all TEXT children of the given node in one string. Between
     * the node values new lines are inserted.
     */
    public String getContent(Element node) {
      NodeList       list;
      Node           item;
      int            i;
      String         result;
      
      result = "";
      list   = node.getChildNodes();
      
      for (i = 0; i < list.getLength(); i++) {
         item = list.item(i);
         if (item.getNodeType() == Node.TEXT_NODE)
            result += "\n" + item.getNodeValue();
      }
         
      return result;
    }


	/** buildInstances parses the BIF document and creates a Bayes Net with its 
	 * nodes specified, but leaves the network structure and probability tables empty.
	 * @param doc: DOM document containing BIF document in DOM tree
	 * @param name: default name to give to the Bayes Net. Will be overridden if specified in the BIF document.
	 */
	void buildInstances(Document doc, String sName) throws Exception {
		NodeList nodelist;
        // Get the name of the network
        nodelist = selectAllNames(doc);
        if (nodelist.getLength() > 0) {
        	sName = ((CharacterData) (nodelist.item(0).getFirstChild())).getData();
        }

        // Process variables
        nodelist = selectAllVariables(doc);
		int nNodes = nodelist.getLength();
		// initialize structure
		FastVector attInfo = new FastVector(nNodes);

        // Initialize
        m_nPositionX = new int[nodelist.getLength()];
        m_nPositionY = new int[nodelist.getLength()];

        // Process variables
        for (int iNode = 0; iNode < nodelist.getLength(); iNode++) {
            // Get element
			FastVector valueslist;
	        // Get the name of the network
    	    valueslist = selectOutCome(nodelist.item(iNode));

			int nValues = valueslist.size();
			// generate value strings
	        FastVector nomStrings = new FastVector(nValues + 1);
	        for (int iValue = 0; iValue < nValues; iValue++) {
	        	Node node = ((Node) valueslist.elementAt(iValue)).getFirstChild();
	        	String sValue = ((CharacterData) (node)).getData();
	        	if (sValue == null) {
	        		sValue = "Value" + (iValue + 1);
	        	}
				nomStrings.addElement(sValue);
	        }
			FastVector nodelist2;
	        // Get the name of the network
    	    nodelist2 = selectName(nodelist.item(iNode));
    	    if (nodelist2.size() == 0) {
    	    	throw new Exception ("No name specified for variable");
    	    }
    	    String sNodeName = ((CharacterData) (((Node) nodelist2.elementAt(0)).getFirstChild())).getData();

			weka.core.Attribute att = new weka.core.Attribute(sNodeName, nomStrings);
			attInfo.addElement(att);

    	    valueslist = selectProperty(nodelist.item(iNode));
			nValues = valueslist.size();
			// generate value strings
	        for (int iValue = 0; iValue < nValues; iValue++) {
                // parsing for strings of the form "position = (73, 165)"
	        	Node node = ((Node)valueslist.elementAt(iValue)).getFirstChild();
	        	String sValue = ((CharacterData) (node)).getData();
                if (sValue.startsWith("position")) {
                    int i0 = sValue.indexOf('(');
                    int i1 = sValue.indexOf(',');
                    int i2 = sValue.indexOf(')');
                    String sX = sValue.substring(i0 + 1, i1).trim();
                    String sY = sValue.substring(i1 + 1, i2).trim();
                    try {
                    	m_nPositionX[iNode] = (int) Integer.parseInt(sX);
                    	m_nPositionY[iNode] = (int) Integer.parseInt(sY);