package com.java.html;import java.io.BufferedReader;import java.io.BufferedWriter;import java.io.File;import java.io.FileNotFoundException;import java.io.FileReader;import java.io.FileWriter;import java.io.IOException;import java.io.RandomAccessFile;import java.util.ArrayList;import java.util.Arrays;import java.util.Collections;import java.util.Comparator;import java.util.HashMap;import java.util.Iterator;import java.util.LinkedList;import java.util.List;import java.util.Map;import com.java.checksql.Check;/** * 相似度对象 */class Simily {private float cos[][];int docNum;public Simily(int docNum) {this.docNum = docNum;cos = new float[docNum + 1][];for (int i = 1; i <= docNum; i++) {cos[i] = new float[docNum – i];// 数组下标j对应的真实文章id为i+j+1；}}/** * 得到指定下标的相似度 */public float getEle(int i, int j) {return cos[i][j];}/** * 打印相似度到 file文件 */public void printSim(String file) throws IOException {BufferedWriter out = new BufferedWriter(new FileWriter(new File(file)));for (int i = 1; i <= docNum; i++) {out.write(i + "");for (int j = i + 1; j <= docNum; j++) {if (cos[i][j – i – 1] != 0) {out.write(" ");out.write(j + ":" + Math.round(cos[i][j – i – 1] * 10000)/ 10000.0);}}out.newLine();out.flush();}out.close();}/** * 得到某一维的长度 */public int getLen(int i) {return cos[i].length;}/** * 更新下标值为 i，j的值 */public void update(int i, int j, float value) {cos[i][j] += value;}public void readSimFile(String file) throws IOException {BufferedReader r = new BufferedReader(new FileReader(file));String line = "";int count = 1;while ((line = r.readLine()) != null) {String[] row = line.split("");if (!row[0].matches("\\d+"))// 检查行号{System.out.println("您输入的格式有问题，在第" + count + "行！");System.exit(1);}int rowID = Integer.parseInt(row[0]);for (int i = 1; i < row.length; i++) {String ele[] = row[i].split(":");// 检查文件号及其相似度int columID=0;float sim=0;if ((!ele[0].matches("\\d+") || (columID=Integer.parseInt(ele[0])) <= 0)&& (!ele[1].matches("(\\d)|(\\d\\.\\d+)") || ((sim=Float.parseFloat(ele[1])) < 0 || sim > 1))) {System.out.println("您输入的格式有问题，在第" + count + "行！");System.exit(1);}if(rowID<columID)cos[rowID][columID – rowID – 1] = sim;elsecos[columID][rowID – columID – 1] = sim;}count++;}r.close();}}class Cluster {public int center;List<Integer> child = new ArrayList<Integer>();public Cluster(int cen) {center = cen;child.add(center);}/** * 为该聚类添加孩子 */public void add(int child) {this.child.add(child);}public void printClu() {System.out.println("聚类中心： " + center);System.out.print("包含的孩子：");for (int i = 0; i < child.size(); i++)System.out.print(child.get(i) + " ");System.out.println();}/** * 判断某个文章是否在此聚类中 */public boolean isIn(int doc) {if (child.contains(doc))return true;return false;}}class DocArgs {public int id;// 文章编号public int p;// 文章的密度public float q = 1;// 密度比此文章大 且距离最小的距离public int qID;// 产生最小距离的文章的IDpublic boolean hasClu = false;// 标志位该文章是否已经有归属,聚类中心的初始值为truepublic boolean isCen = false;// 判断是否为聚类中心// 存放孩子 构造以聚类中心的森林 方便寻找 该聚类包含的所有文章List<Integer> child = new LinkedList<Integer>();public DocArgs(int id) {this.id = id;}}public class FastCluster {// 文章数int docNum;// 存放两两文章相似度（非零元素上三角）Simily cos = null;// 存放文章的属性DocArgs[] doc;// 用于存放生成的簇List<Cluster> clu = new ArrayList<Cluster>();public FastCluster(String file) throws IOException {System.out.println("开始读入相似度文件……");getDocNum(file);cos = new Simily(docNum);readSim(file);System.out.println("相似度文件已经读入……");}/** * 得到文章数量 * */private void getDocNum(String file) throws IOException {long len = 0;String line = "";RandomAccessFile raf = null;try {raf = new RandomAccessFile(file, "r");len = raf.length();long pos = len – 2;while (pos >= 0) {raf.seek(pos);if (raf.readByte() == '\n') {line = raf.readLine();;break;}pos–;}} catch (IOException e) {e.printStackTrace();}String a[] = line.split("\t", 2);this.docNum = Integer.parseInt(a[0]);}/** * 输出相似度到磁盘 * */public void printCOS(String file) throws IOException {cos.printSim(file);}public void readSim(String file) throws IOException {cos.readSimFile(file);}/** * 得到聚类的中心 d为指定的两两文章的距离 * * @throws IOException */public void cluster(float d, float pq, String pqFile) throws IOException {initDoc();System.out.println("开始获得每个文章的密度P……");getP(d);System.out.println("开始获得每个文章的q……");getQ();System.out.println("寻找聚类中心……");getCenter(pq);System.out.println("对文章进行分类……");divideDoc();// 分类finalClu();// 为每一个中心分配文章System.out.println("分类完毕……");System.out.println("开始打印pq文件……");printPQ(pqFile);System.out.println("pq文件打印结束……");}private void printPQ(String pqFile) throws IOException {BufferedWriter out = new BufferedWriter(new FileWriter(pqFile));for (int i = 1; i <= docNum; i++) {out.write(doc[i].id + "" + doc[i].p + "" + doc[i].q + ""+ doc[i].q * doc[i].p);out.newLine();}out.close();}/** * 从控制台输出 各个聚类及其所包含的文件 */public void printClu() {for (int i = 0; i < clu.size(); i++) {clu.get(i).printClu();}}public void printFile(String file) throws IOException {BufferedWriter out = new BufferedWriter(new FileWriter(file));for (int i = 1; i <= docNum; i++) {int cen = 0;for (int j = 0; j < clu.size(); j++)if (clu.get(j).isIn(i)) {cen = clu.get(j).center;break;}out.write(i + "" + cen);out.newLine();out.flush();}out.close();}/** * 对doc初始化 */private void initDoc() {doc = new DocArgs[docNum + 1];for (int i = 0; i <= docNum; i++)doc[i] = new DocArgs(i);doc[0].p = -100;}/** * 找到聚类的中心点,初始化每个聚类的簇 * */public void getCenter(float pq) {for (int i = 1; i <= docNum; i++) {if (doc[i].p * doc[i].q >= pq) {doc[i].isCen = true;doc[i].hasClu = true;clu.add(new Cluster(doc[i].id));}}}/** * 得到聚类的密度 */private void getP(float d) {for (int doc1 = 1; doc1 < docNum; doc1++) {for (int j = 0; j < cos.getLen(doc1); j++) {int doc2 = j + doc1 + 1;float cosn = cos.getEle(doc1, j);if (cosn >= d) {doc[doc1].id = doc1;doc[doc1].p++;doc[doc2].id = doc2;doc[doc2].p++;}}}}/** * 为了减少时间要对所有的文章按密度递增的顺序排序； */private void pSort() {Arrays.sort(doc, new Comparator<DocArgs>() {@Overridepublic int compare(DocArgs o1, DocArgs o2) {if (o1.p > o2.p)return 1;else if (o1.p < o2.p)return -1;elsereturn 0;}});}/** * 为了在寻找聚类中心递归运算时候，方便查找要对所有的文章按id递增的顺序排序； */private void IDSort() {Arrays.sort(doc, new Comparator<DocArgs>() {@Overridepublic int compare(DocArgs o1, DocArgs o2) {if (o1.id > o2.id)return 1;else if (o1.id < o2.id)return -1;elsereturn 0;}});}/** * 得到文章的第二个属性q：比此文章大的，且距离最近的距离 */private void getQ() {pSort();// for(int i=0;i<=docNum;i++)// System.out.println(doc[i].id+" "+doc[i].p+" "+doc[i].q+" "+doc[i].qID);for (int i = 1; i <= docNum; i++) {int j = i + 1;// 排除密度和它相等的文章while (j <= docNum && doc[i].p == doc[j].p)j++;for (; j <= docNum; j++) {int min = doc[i].id > doc[j].id ? doc[j].id : doc[i].id;int max = doc[i].id < doc[j].id ? doc[j].id : doc[i].id;if (doc[i].q > 1 – cos.getEle(min, max – min – 1))// 更新距离{doc[i].q = 1 – cos.getEle(min, max – min – 1);doc[i].qID = doc[j].id;}}}}/** * 为每个聚类中心分配 文章 递归算法 */private void divideDoc() {IDSort();// for(int i=0;i<=docNum;i++)// System.out.println(doc[i].id+" "+doc[i].p+" "+doc[i].q+" "+doc[i].qID);for (int i = 1; i <= docNum; i++)if (doc[i].hasClu == false)divide(i);}private void divide(int id) {if (doc[id].hasClu == false) {doc[id].hasClu = true;doc[doc[id].qID].child.add(id);divide(doc[id].qID);}}/** * 递归得到cen聚类中心拥有的文章 */public void finalClu() {for (int i = 0; i < clu.size(); i++) {Cluster cluster = clu.get(i);int cen = cluster.center;getSubCluster(cen, cluster.child);}}public void getSubCluster(int cen, List<Integer> clu) {for (int i = 0; i < doc[cen].child.size(); i++) {int center = doc[cen].child.get(i);clu.add(center);getSubCluster(center, clu);}}private static boolean checkArgs(String args[]) {if (args.length != 5) {System.out.println("您输入的参数个数不正确，请确保是5个！！");return false;}if (!new File(args[0]).exists()) {System.out.println("您输入文件不存在，请仔细检查！！");return false;}if (!(args[3].matches("(\\d)|(\\d\\.\\d+)"))|| (Float.parseFloat(args[3]) < 0 || Float.parseFloat(args[3]) > 1)) {System.out.println("您输入参数d不合法，请仔细检查！！");return false;}if (!(args[4].matches("(\\d+\\.\\d+)|(\\d+)"))|| (Float.parseFloat(args[4]) < 0)) {System.out.println("您输入参数pq不合法，请仔细检查！！");return false;}return true;}/* * args数组的参数：args[0]:为相似度文件路径 args[1]:为pq文件的输出路径 args[2]:为聚类文件的输出路径 * args[3]:为参数为相似度 args[4]:为参数pq */public static void main(String[] args) throws IOException {System.out.println("开始运行……");System.out.println("正在检查参数……");if (!checkArgs(args))System.exit(1);System.out.println("参数正确……");FastCluster fc = new FastCluster(args[0]);fc.cluster(Float.parseFloat(args[3]), Float.parseFloat(args[4]),args[1]);System.out.println("开始打印聚类文件……");fc.printFile(args[2]);System.out.println("运行结束……");// D:\\世界杯\\sim.txt// D:\\世界杯\\cosqq.txt// D:\\世界杯\\qq.txt}}
，你被雨淋湿的心，是否依旧。
