id3.java

id3算法代码,用C 语言实现的.小弟急需JAVA语言写的ID3算法代码,麻烦有的大哥大姐共享下,

	 * @throws IOException
	 */
	private Matrix buildTestMatrix(File testFile,Matrix testData) throws IOException
	{
		FileInputStream fin=new FileInputStream(testFile);
		BufferedReader reader=new BufferedReader(new InputStreamReader(fin));
		String line=reader.readLine();
		ArrayList<RowData> rows=testData.matrix;
		while(line!=null)
		{
			RowData rowdata=new RowData();
			String[] items=line.split(" ");
			ArrayList<String> row=rowdata.rowdata;
			for(int i=0;i<items.length;i++)
				row.add(items[i]);
			rows.add(rowdata);
			line=reader.readLine();
		
		}
		reader.close();
		return testData;
		
	}
	
	
	/**
	 * 由规则列表对测试数据集进行分类
	 * @param testData 测试数据集
	 * @param ruleList 规则列表
	 * @return 带有类标识的数据集
	 */
	private Matrix genTestMatrix(Matrix testData,ArrayList<String> ruleList)
	{
		ArrayList<RowData> rows=testData.matrix;
		for(int i=0;i<rows.size();i++)
		{
			ArrayList<String> row=rows.get(i).rowdata;
			String target="";
			//对每个规则,检验是否匹配
			for(int j=0;j<ruleList.size();j++)
			{
				String rule=ruleList.get(j);
				int labelIndex=rule.indexOf(":");
				while(labelIndex!=-1)
				{
					//冒号前的字符表示主属性的下标
					char carrayIndex=rule.charAt(labelIndex-1);
					//冒号后的字符表示主属性下的子属性标识
					String label=""+rule.charAt(labelIndex+1);
					int arrayIndex=Integer.parseInt(""+carrayIndex);
					if(row.get(arrayIndex).equals(label))
					{
						rule=rule.substring(labelIndex+1);
						labelIndex=rule.indexOf(":");
					}
					else
					{
						break;
					}
				}
				//成功的匹配
				if(labelIndex==-1)
				{
					target+=rule.charAt(rule.length()-1);
					break;
				}
			} //end for(int j=0;i<ruleList
			//将类标识添加到测试数据集末尾
			row.add(target);
		}
		return testData;
	}
	
	
	/**
	 * 测试分类的正确性
	 * @param testData 测试数据集
	 * @param testClassFile 目标类测试文件
	 * @throws IOException
	 */
	private void checkClass(Matrix testData,File testClassFile) throws IOException
	{
		FileInputStream fin=new FileInputStream(testClassFile);
		BufferedReader reader=new BufferedReader(new InputStreamReader(fin));
		String line=reader.readLine();
		ArrayList<String> classList=new ArrayList<String>();
		int totalCounts=0; //数据行总数
		int correctCounts=0; //正确标识的行数
		while(line!=null)
		{
			totalCounts++;
			classList.add(line);
			line=reader.readLine();
		}
		reader.close();
		ArrayList<RowData> rows=testData.matrix;
		for(int i=0;i<rows.size();i++)
		{
			ArrayList<String> row=rows.get(i).rowdata;
			String cLabel=row.get(row.size()-1);
			if(cLabel.equals(classList.get(i)))
				correctCounts++;
		}
		float correctPercent=(float)correctCounts/totalCounts;
		System.out.println("分类正确率:"+correctPercent*100+"%");
	}
	
	
	/**
	 * 测试分类正确性(外部接口)
	 * @param testFile
	 * @param testClassFile
	 * @throws IOException
	 */
	public void checkClassification(File testFile,File testClassFile) throws IOException
	{
		testData=buildTestMatrix(testFile, testData);
		testData=genTestMatrix(testData, ruleList);
		this.checkClass(testData, testClassFile);
	}
	
	private AttributesList attributeList;      //属性列表
	private Matrix dataSet;                    //训练数据集
	private Matrix testData;                   //测试数据集
	private ArrayList<String> ruleList;        //规则列表
	private final double LN2=Math.log(2.0);   //ln(2)
	Node root;                                  //根节点
}

/**
 * 判定树节点类
 * @author Lam
 *
 */
class Node
{
	Node()
	{
		this("",-1,0);
	}
	
	/**
	 * 
	 * @param theLabel 节点标识
	 * @param theDecision 是否叶节点
	 * @param childCounts 分支数目
	 */
	Node(String theLabel,int theDecision,int childCounts)
	{
		label=theLabel;
		decision=theDecision;
		if(childCounts>0)
		{
			childs=new Node[childCounts];
		}
	}
	
	/**
	 * 
	 * @param theLabel 节点标识
	 * @param thePath  节点路径
	 * @param theDecision 是否叶节点
	 * @param childCounts 分支数目
	 */
	Node(String theLabel,String thePath,int theDecision,int childCounts)
	{
		this.label=theLabel;
		this.path=thePath;
		this.decision=theDecision;
		if(childCounts>0)
		{
			childs=new Node[childCounts];
		}
	}
	
	String label; //节点标识
	String path; //节点路径
	int decision; //区分节点标志,-1是普通节点,1是叶节点
	Node[] childs; //分支
}

/**
 * 数据集类
 * @author Lam
 *
 */
class Matrix
{
	Matrix()
	{
		matrix=new ArrayList<RowData>();
	}
	
	void setWidth(int theWidth)
	{
		width=theWidth;
	}
	
	int width;
	ArrayList<RowData> matrix;
}

/**
 * 行数据
 * @author Lam
 *
 */
class RowData
{
	RowData()
	{
		rowdata=new ArrayList<String>();
	}
	ArrayList<String> rowdata;
}

/**
 * 属性列表
 * @author Lam
 *
 */
class AttributesList
{
	AttributesList()
	{
		attributes=new ArrayList<Column>();
		rowcounts=0;
		inentropy=0.0;
	}
	int rowcounts;  //数据集行数
	double inentropy; //类区分整个数据集的熵
	ArrayList<Column> attributes;
	
}

/**
 * 属性列
 * @author Lam
 *
 */
class Column
{
	Column()
	{
		attributes=new ArrayList<Attribute>();
		label="";
	}
	ArrayList<Attribute> attributes; //列下的属性列表
	String label; //列标识
}

/**
 * 主属性
 * @author Lam
 *
 */
class Attribute
{
	Attribute()
	{
		targets=new ArrayList<Target>();
	}
	
	int counts; //属性包含元素个数
	String label; //属性标识
	ArrayList<Target> targets; //属性下的目标分类
}

/**
 * 目标类
 * @author Lam
 *
 */
class Target
{
	String label; //目标类标识
	int counts;   //目标类包含的元素个数
}

/**
 * 基于某个属性列表的目标类列表
 * 用于计算对于一个样本分类所需的信息I
 * @author Lam
 *
 */
class TotalTarget
{
	TotalTarget()
	{
		totalTargets=new ArrayList<Target>();
		totalCount=0;
	}
	ArrayList<Target> totalTargets; //目标类列表
	int totalCount;  //总的元素个数
}

public class ID3 {

	
	public ID3()
	{
		//dataSet=new Matrix();
		//attrList=new AttributesList();
		//totalTarget=new ArrayList<Target>();
	}
	/**
	 * 从数据文件中构造数据表Matrix
	 * @param dataFile 数据文件
	 * @param attrList 属性列表
	 * @return
	 * @throws IOException
	 */
	static Matrix buildMatrix(File dataFile,AttributesList attrList) throws IOException
	{
		FileInputStream fin=new FileInputStream(dataFile);
		BufferedReader reader=new BufferedReader(new InputStreamReader(fin));
		String line=reader.readLine();
		String[] items;
		Matrix matrixD=new Matrix();
		RowData rowD;
		items=line.split(" ");
		//属性列表初始化
		for(int i=0;i<items.length;i++)
		{
			Column col=new Column();
			col.label="Attribute"+i;
			attrList.attributes.add(col);
			
		}
		matrixD.setWidth(items.length);
		while(line!=null)
		{
			items=line.split(" ");
			rowD=new RowData();
			for(int i=0;i<items.length;i++)
			{
				
				rowD.rowdata.add(items[i]);
			}
			matrixD.matrix.add(rowD);
			line=reader.readLine();
			
		}
		reader.close();
		return matrixD;
	}
	
	
	/**
	 * 读取类文件以及Matrix数据表
	 * 构造属性表
	 * @param classFile 类文件
	 * @param theData   数据表
	 * @throws IOException
	 */
	static void buildAttribute(File classFile,Matrix theData,AttributesList attrList) throws IOException
	{
		FileInputStream fin=new FileInputStream(classFile);
		BufferedReader reader=new BufferedReader(new InputStreamReader(fin));
		String line=reader.readLine();
		//AttributesList attrList=new AttributesList();
		//类
		Target ta;
		TotalTarget totalTargets=new TotalTarget();
		ArrayList<Target> totalList=totalTargets.totalTargets;
		//列集(每列是一个主属性,包含若干子属性)
		ArrayList<Column> cols=attrList.attributes;
		//行数
		int lncount=0;
		while(line!=null)
		{
			//*******目标各属性的总个数
			int t=0;
			for(;t<totalList.size();t++)
			{
				Target tag=totalList.get(t);
				if(tag.label.equals(line))
				{
					totalTargets.totalCount++;
					tag.counts++;
					break;
				}
			}
			//类标识尚未建立
			if(t==totalList.size())
			{
				ta=new Target();
				ta.label=line;
				ta.counts++;
				totalList.add(ta);
				totalTargets.totalCount++;
			}
			//*******//
			//*******读取Matrix的每一行,构造Attribute
			ArrayList<RowData> rows=theData.matrix;
			//
			ArrayList<String> rowdata=rows.get(lncount++).rowdata;
			//row的长度和cols的长度一样
			for(int i=0;i<cols.size();i++)
			{
				int j=0;
				//取得每一列的引用
				Column col=cols.get(i);
				//列的子属性列表
				ArrayList<Attribute> attCol=col.attributes;
				
				
				for( ;j<attCol.size();j++)
				{
					
					//列的第j个属性
					Attribute attr=attCol.get(j);
					
					//
					if(attr.label.equals(rowdata.get(i)))
					{
						attr.counts++;
						break;
					}
				} //end for( ;j<attCol.size
				//这个属性尚未构建
				if(j==attCol.size())
				{
					Attribute attr=new Attribute();
					attr.label=rowdata.get(i);
					attr.counts++;
					//设置子属性的目标类个数
					ArrayList<Target> tarList=attr.targets;
					Target target=new Target();
					target.label=line;
					target.counts++;
					tarList.add(target);
					
					attCol.add(attr);
				}
				else
				{
					//找到这个属性,对属性下的子属性的类标识计数
					int k=0;
					Attribute attr=attCol.get(j);
					ArrayList<Target> tarList=attr.targets;
					for(;k<tarList.size();k++)
					{
						Target tar=tarList.get(k);
						if(tar.label.equals(line))
						{
							tar.counts++;
							break;
						}
					}
					//在Attribute表中找不到这个目标属性
					//Attribute表为空时
					if(k==tarList.size())
					{
						Target newtar=new Target();
						newtar.label=line;
						newtar.counts++;
						tarList.add(newtar);
					}
				}
				
			}
//			将类标识加入到Matrix中
			rowdata.add(line);
			//读取下一行
			line=reader.readLine();
			
		}
		reader.close();
		//加入类标识,列数加1
		theData.setWidth(theData.width+1);
		//计算区分整个数据集的熵
		int totalCounts=totalTargets.totalCount;
		double inentropy=0.0;
		for(int i=0;i<totalList.size();i++)
		{
			Target tar=totalList.get(i);
			int pCount=tar.counts;
			double p=(double)pCount/totalCounts;
			inentropy+=p*Math.log(p)/LN2;
		}
		inentropy=-1*inentropy;
		attrList.inentropy=inentropy;
		attrList.rowcounts=theData.matrix.size();
		//return attrList;
	}
	
	private static final double LN2=Math.log(2.0); //ln(2)
	
	
	/**
	 * 数据集测试
	 * @param args
	 */
	public static void main(String[] args)throws IOException{
		// TODO 自动生成方法存根

		String dataS="Train_Attr.txt";
		File dataFile=new File(dataS);
		if(!dataFile.exists())
		{
			System.err.println("在当前目录下找不到文件Train_Attr.txt!");
			System.exit(-1);
		}
		String classS="Train_Class.txt";
		File classFile=new File(classS);
		if(!classFile.exists())
		{
			System.err.println("在当前目录下找不到文件Train_Class.txt!");
			System.exit(-1);
		}
		String testDataS="Test_Attr.txt";
		File testDataFile=new File(testDataS);
		if(!testDataFile.exists())
		{
			System.err.println("在当前目录下找不到文件Test_Attr.txt!");
			System.exit(-1);
		}
		String testClassS="Test_Class.txt";
		File testClassFile=new File(testClassS);
		if(!testClassFile.exists())
		{
			System.err.println("在当前目录下找不到文件Test_Class.txt!");
			System.exit(-1);
		}
		ID3Tree id3tree=new ID3Tree();
		System.out.println("开始构建判定树...");
		id3tree.start(dataFile, classFile);
		//id3tree.start(testDataFile, testClassFile);
		System.out.println("由训练数据集构建判定树完成.以下是导出的自定义规则:");
		//id3tree.printTree();
		id3tree.showRules();
		System.out.println("测试分类正确性...");
		id3tree.checkClassification(testDataFile, testClassFile);
		
	}

}