softwarerec/src/main/java/com/ow2/rec/util/SimilarityCounter.java

//Copyright (c) 2016 Yarong Zeng, NUDT.
package com.ow2.rec.util;

import java.util.HashSet;
import java.util.LinkedList;
import java.util.List;
import java.util.Set;


public class SimilarityCounter {
	
	
	//两个文本结合 以及各自个各个分量的权重的倒数
	public static float countSimilarity(List<String>prj,int[]prjWeight,List<String>tag,int[]tagWeight){
		
		//如果没有共同的元素就直接返回0
		List<String> prjTmp = new LinkedList<String>(prj);
		prjTmp.retainAll(tag);
		if(prjTmp.size() == 0){
			return 0;
		}
		
		//如果完全匹配就直接返回1
		if(prj.containsAll(tag) && tag.containsAll(prj))
			return 1;
		
		//如果项目名称为单个字母，直接返回0
		if(prj.size() == 1 && prj.get(0).length() == 1){
			return 0;
		}
		
		
		Set<String> union= new HashSet<String>();//此集合用于存储项目和标签匹配时他们的原子性元素的并集
		union.addAll(prj);
		union.addAll(tag);
		
		//生成两个向量
		float prjVector[] = new float[union.size()];
		float tagVector[] = new float[union.size()];
		
		int i = 0;
		for(String item : union){
			if(prj.contains(item))
				prjVector[i] = 1.0F / prjWeight[prj.indexOf(item)];
			
			if(tag.contains(item))
				tagVector[i] = 1.0F / tagWeight[tag.indexOf(item)];
			i++;
		}
		
		//两个向量的点积
		float dotMulti = 0,absPrjUnSqrt = 0, absTagUnSqrt  =0;
		
		for(i = 0; i< prjVector.length; i++){
			dotMulti += prjVector[i] * tagVector[i];
			absPrjUnSqrt  +=  prjVector[i] * prjVector[i];
			absTagUnSqrt += tagVector[i] * tagVector[i];
		}
		
		
		
		float absPrj = 0, absTag =0;
		absPrj = (float) Math.sqrt(absPrjUnSqrt );
		absTag = (float) Math.sqrt(absTagUnSqrt );
		
		return formatFloat(dotMulti / ( absPrj * absTag),3);
	}
	
	private static float formatFloat(float f,int num){
		int base = (int) Math.pow(10,num);		
		return (float) (Math.round(f * base))/base;
	}
}