代码如下：package listNet_xiuzhang;import java.io.BufferedReader;import java.io.File;import java.io.FileInputStream;import java.io.FileWriter;import java.io.InputStreamReader;public class listNet {//文件总行数(标记数)private static int sumLabel;//特征值 46个 (标号1-46)private static double feature[][] = new double[100000][48];//特征值权重 46个 (标号1-46)private static double weight [] = new double[48];//相关度 其值有0-2三个级别 从1开始记录private static int label [] = new int[1000000];//查询id 从1开始记录private static int qid [] = new int[1000000];//每个Qid的doc数量private static int doc_ofQid[] = new int[100000]; private static int ITER_NUM=30;//迭代次数private static int weidu=46;//特征数private static int qid_Num=0;//Qid数量private static int tempQid=-1;//临时Qid数private static int tempDoc=0;//临时doc数/*** 函数功能 读取文件 * 参数 String filePath 文件路径 */public static void ReadTxtFile(String filePath) {try {String encoding="GBK";File file=new File(filePath);if(file.isFile() && file.exists()) { //判断文件是否存在InputStreamReader read = new InputStreamReader(new FileInputStream(file), encoding);BufferedReader bufferedReader = new BufferedReader(read);String lineTxt = null;sumLabel =1; //初始化从1记录//按行读取数据并分解数据while((lineTxt = bufferedReader.readLine()) != null) {String str = null;int lengthLine = lineTxt.length();//获取数据 字符串空格分隔String arrays[] = lineTxt.split(" ");for(int i=0; i<arrays.length; i++) {//获取每行样本的Label值if(i==0) {label[sumLabel] = Integer.parseInt(arrays[0]);}else if(i>=weidu+2){ //读取至#跳出 0-label 1-qid 2:47-特征continue;}else {String subArrays[] = arrays[i].split(":"); //特征:特征值if(i==1) { //获取qid//判断是否是新的Qidif(tempQid != Integer.parseInt(subArrays[1])) {if(tempQid != -1){ //不是第一次出现新Qid//赋值上一个为qid_Num对应的tempDoc个文档doc_ofQid[qid_Num]=tempDoc;tempDoc=0;}//当tempQid不等于当前qid时下标加1//相等则直接跳至Doc加1直到不等qid_Num++;tempQid=Integer.parseInt(subArrays[1]);}tempDoc++; //新的文档qid[sumLabel] = Integer.parseInt(subArrays[1]);}else { //获取46维特征值int number = Integer.parseInt(subArrays[0]); //判断特征double value = Double.parseDouble(subArrays[1]);feature[sumLabel][number] = value; //number数组标号:1-46}}}sumLabel++;}doc_ofQid[qid_Num]=tempDoc;read.close();} else {System.out.println("找不到指定的文件\n");}} catch (Exception e) {System.out.println("读取文件内容出错");e.printStackTrace();}}/** * 学习排序 * 训练模型得到46维权重 */public static void LearningToRank() {//变量double index [] = new double[1000000];double tao [] = new double[1000000];double yita=0.00003;//初始化for(int i=0;i<weidu+2;i++) { //从1到136为权重，，0和137无用weight[i] = (double) 1.0; //权重初值}System.out.println("training…");//计算权重 学习算法for(int iter = 0; iter<ITER_NUM; iter++) //迭代ITER_NUM次{System.out.println("—迭代次数:"+iter);int now_doc=0; //全局文档索引for(int i=1; i<=qid_Num; i++) //总样qid数 相当于两层循环T和m{double delta_w[] = new double[weidu+2]; //46个梯度组成的向量int doc_of_i=doc_ofQid[i]; //该Qid的文档数//得分f(w),一个QID有多个文档，一个文档为一个分,所以一个i对应一个分数数组double fw[] = new double[doc_of_i+2];/* 第一步 算得分数组fw fin */for(int k=1;k<=doc_of_i;k++) { //初始化fw[k]=0.0;}for(int k=1;k<=doc_of_i;k++) { //每个文档的得分for(int p=1;p<=weidu;p++) {fw[k]=fw[k]+weight[p]*feature[now_doc+k][p]; //算出这个文档的分数}}/** 第二步 算梯度delta_w向量* a=Σp*x,a是向量* b=Σexpf(x),b是数字* c=expf(x)*x,c是向量* 最终结果delta_w是向量*/double[] a=new double[weidu+2],c=new double[weidu+2];for(int k=0;k<weidu+2;k++){a[k]=0.0;} //初始化for(int k=0;k<weidu+2;k++){c[k]=0.0;} //初始化double b=0.0;//算a：—-for(int k=1; k<=doc_of_i; k++) {double p=1.0; //先不topKdouble[] temp=new double[48];for(int q=1;q<=weidu;q++) {//算P: —-第q个向量排XX的概率是多少//分母：double fenmu=0.0;for(int m=1;m<=doc_of_i;m++) {fenmu=fenmu+Math.exp(fw[m]); //所有文档得分}//top-1 exp(s1) / exp(s1)+exp(s2)+..+exp(sn)for(int m=1;m<=doc_of_i;m++) {p=p*(Math.exp(fw[m])/fenmu);}//算积temp[q]=temp[q]+p*feature[now_doc+k][q];}for(int q=1; q<=weidu; q++){a[q]=a[q]+temp[q];}} //End a//算b：—- fin.for(int k=1; k<=doc_of_i; k++){b=b+Math.exp(fw[k]);}//算c：—-for(int k=1; k<=doc_of_i; k++){double[] temp=new double[weidu+2];for(int q=1; q<=weidu; q++){temp[q]=temp[q]+Math.exp(fw[k])*feature[now_doc+k][q];}for(int q=1; q<=weidu; q++){c[q]=c[q]+temp[q];}}//算梯度：delta_x=-a+1/b*cfor(int q=1; q<=weidu; q++){delta_w[q]= (-1)*a[q] + ((1.0/b)*c[q]);}//**********/* 第三步 更新权重 fin. */for(int k=1; k<=weidu; k++){weight[k]=weight[k]-yita*delta_w[k];}now_doc=now_doc+doc_of_i; //更新当前文档索引}} //End 迭代次数//输出权重for(int i=1;i<=weidu;i++) //从1到136为权重，0和137无用{System.out.println(i+"wei:"+weight[i]);}}/** * 输出权重到文件fileModel * @param fileModel */public static void WriteFileModel(String fileModel) {//输出权重到文件try {System.out.println("write start.总行数："+sumLabel);FileWriter fileWriter = new FileWriter(fileModel);//写数据fileWriter.write("## ListNet");fileWriter.write("\r\n");fileWriter.write("## Epochs = "+ITER_NUM);fileWriter.write("\r\n");fileWriter.write("## No. of features = 46");fileWriter.write("\r\n");fileWriter.write("1 2 3 4 5 6 7 8 9 10 … 39 40 41 42 43 44 45 46");fileWriter.write("\r\n");fileWriter.write("0");fileWriter.write("\r\n");for(int k=0; k<weidu; k++){fileWriter.write("0 "+k+" "+weight[k+1]);fileWriter.write("\r\n");}fileWriter.close();System.out.println("write fin.");} catch(Exception e) {System.out.println("写文件内容出错");e.printStackTrace();}}/** * 预测排序 * 正规应对test.txt文件进行打分排序 * 但我们是在Hadoop实现该打分排序步骤 此函数仅测试train.txt打分 */public static void PredictRank(String fileScore) {//输出得分try {System.out.println("write start.总行数："+sumLabel);String encoding = "GBK";FileWriter fileWriter = new FileWriter(fileScore);//写数据for(int k=1; k<sumLabel; k++){double score=0.0;for(int j=1;j<=weidu;j++){score=score+weight[j]*feature[k][j];}fileWriter.write("qid:"+qid[k]+" score:"+score+" label:"+label[k]);fileWriter.write("\r\n");}fileWriter.close();System.out.println("write fin.");} catch(Exception e) {System.out.println("写文件内容出错");e.printStackTrace();}}/** * 主函数 */public static void main(String args[]) {String fileInput = "Fold1\\train.txt";//训练String fileModel = "model_weight.txt";//输出权重模型String fileScore = "score_listNet.txt";//输出得分//第1步 读取文件并解析数据System.out.println("read…");ReadTxtFile(fileInput);System.out.println("read and write well.");//第2步 排序计算LearningToRank();//第3步 输出模型WriteFileModel(fileModel);//第4步 打分预测排序PredictRank(fileScore); }/* * End */}
因为它抚平了心底的不安;当你尝到了极品的美食，
