聚类算法之KMeans(Java实现)

聚类算法之KMeans(Java实现)

KMeans算法的基本思想是初始随机给定K个簇中心，按照最邻近原则把待分类样本点分到各个簇。然后按平均法重新计算各个簇的质心，从而确定新的簇心。一直迭代，直到簇心的移动距离小于某个给定的值。

当初始簇心选行不好时，KMeans的结果会很差，所以一般是多运行几次，按照一定标准（比如簇内的方差最小化）选择一个比较好的结果。

下图给出对坐标点的聚类结果：

下面给出核心算法的代码：

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173

/**   * Author: Orisun   * Date: Sep 10, 2011   * FileName: KMeans.java   * Function:   */ package  orisun;   import  java.io.File; import  java.util.ArrayList; import  java.util.Iterator; import  java.util.Random;   public  class  KMeans {        int  k; // 指定划分的簇数      double  mu; // 迭代终止条件，当各个新质心相对于老质心偏移量小于mu时终止迭代      double [][] center; // 上一次各簇质心的位置      int  repeat; // 重复运行次数      double [] crita; // 存放每次运行的满意度        public  KMeans( int  k, double  mu, int  repeat, int  len) {          this .k = k;          this .mu = mu;          this .repeat = repeat;          center = new  double [k][];          for  ( int  i = 0 ; i < k; i++)              center[i] = new  double [len];          crita = new  double [repeat];      }        // 初始化k个质心，每个质心是len维的<a target="_blank" style="color: #0000F0; display:inline; position:static; background:none;" href="http://www.so.com/s?q=%E5%90%91%E9%87%8F&ie=utf-8&src=se_lighten_f">向量</a>，每维均在left--right之间      public  void  initCenter( int  len, ArrayList<DataObject> objects) {          Random random = new  Random(System.currentTimeMillis());          int [] count = new  int [k]; // 记录每个簇有多少个元素          Iterator<DataObject> iter = objects.iterator();          while  (iter.hasNext()) {              DataObject object = iter.next();              int  id = random.nextInt( 10000 )%k;              count[id]++;              for  ( int  i = 0 ; i < len; i++)                  center[id][i] += object.getVector()[i];          }          for  ( int  i = 0 ; i < k; i++) {              for  ( int  j = 0 ; j < len; j++) {                  center[i][j] /= count[i];              }          }      }        // 把数据集中的每个点归到离它最近的那个质心      public  void  classify(ArrayList<DataObject> objects) {          Iterator<DataObject> iter = objects.iterator();          while  (iter.hasNext()) {              DataObject object = iter.next();              double [] vector = object.getVector();              int  len = vector.length;              int  index = 0 ;              double  neardist = Double.MAX_VALUE;              for  ( int  i = 0 ; i < k; i++) {                  // double dist = Global.calEuraDist(vector, center[i], len);                  // //使用<a target="_blank" style="color: #0000F0; display:inline; position:static; background:none;" href="http://www.so.com/s?q=%E6%AC%A7%E6%B0%8F%E8%B7%9D%E7%A6%BB&ie=utf-8&src=se_lighten_f">欧氏距离</a>                  double  dist = Global.calEditDist(vector, center[i], len); // 使用<a target="_blank" style="color: #0000F0; display:inline; position:static; background:none;" href="http://www.so.com/s?q=%E7%BC%96%E8%BE%91%E8%B7%9D%E7%A6%BB&ie=utf-8&src=se_lighten_f">编辑距离</a>                  if  (dist < neardist) {                      neardist = dist;                      index = i;                  }              }              object.setCid(index);          }      }        // 重新计算每个簇的质心，并判断终止条件是否满足，如果不满足更新各簇的质心,如果满足就返回true.len是数据的维数      public  boolean  calNewCenter(ArrayList<DataObject> objects, int  len) {          boolean  end = true ;          int [] count = new  int [k]; // 记录每个簇有多少个元素          double [][] sum = new  double [k][];          for  ( int  i = 0 ; i < k; i++)              sum[i] = new  double [len];          Iterator<DataObject> iter = objects.iterator();          while  (iter.hasNext()) {              DataObject object = iter.next();              int  id = object.getCid();              count[id]++;              for  ( int  i = 0 ; i < len; i++)                  sum[id][i] += object.getVector()[i];          }          for  ( int  i = 0 ; i < k; i++) {              if  (count[i] != 0 ) {                  for  ( int  j = 0 ; j < len; j++) {                      sum[i][j] /= count[i];                  }              }              // 簇中不包含任何点,及时调整质心              else  {                  int  a=(i+ 1 )%k;                  int  b=(i+ 3 )%k;                  int  c=(i+ 5 )%k;                  for  ( int  j = 0 ; j < len; j++) {                      center[i][j] = (center[a][j]+center[b][j]+center[c][j])/ 3 ;                  }              }          }          for  ( int  i = 0 ; i < k; i++) {              // 只要有一个质心需要移动的距离超过了mu，就返回false              // if (Global.calEuraDist(sum[i], center[i], len) >= mu) { //使用欧氏距离              if  (Global.calEditDist(sum[i], center[i], len) >= mu) { // 使用编辑距离                  end = false ;                  break ;              }          }          if  (!end) {              for  ( int  i = 0 ; i < k; i++) {                  for  ( int  j = 0 ; j < len; j++)                      center[i][j] = sum[i][j];              }          }          return  end;      }        // 计算各簇内数据和方差的<a target="_blank" style="color: #0000F0; display:inline; position:static; background:none;" href="http://www.so.com/s?q=%E5%8A%A0%E6%9D%83%E5%B9%B3%E5%9D%87&ie=utf-8&src=se_lighten_f">加权平均</a>，得出本次聚类的满意度.len是数据的维数      public  double  getSati(ArrayList<DataObject> objects, int  len) {          double  satisfy = 0.0 ;          int [] count = new  int [k];          double [] ss = new  double [k];          Iterator<DataObject> iter = objects.iterator();          while  (iter.hasNext()) {              DataObject object = iter.next();              int  id = object.getCid();              count[id]++;              for  ( int  i = 0 ; i < len; i++)                  ss[id] += Math.pow(object.getVector()[i] - center[id][i], 2.0 );          }          for  ( int  i = 0 ; i < k; i++) {              satisfy += count[i] * ss[i];          }          return  satisfy;      }        public  double  run( int  round, DataSource datasource, int  len) {          System.out.println( "第"  + round + "次运行" );          initCenter(len,datasource.objects);          classify(datasource.objects);          while  (!calNewCenter(datasource.objects, len)) {              classify(datasource.objects);          }          datasource.printResult(datasource.objects, k);          double  ss = getSati(datasource.objects, len);          System.out.println( "加权方差："  + ss);          return  ss;      }        public  static  void  main(String[] args) {          DataSource datasource = new  DataSource();          datasource.readMatrix( new  File( "/home/orisun/test/dot.mat" ));          datasource.readRLabel( new  File( "/home/orisun/test/dot.rlabel" )); //      datasource.readMatrix(new File("/home/orisun/text.normalized.mat")); //      datasource.readRLabel(new File("/home/orisun/text.rlabel"));          int  len = datasource.col;          // 划分为6个簇，质心移动小于1E-8时终止迭代，重复运行7次          KMeans km = new  KMeans( 4 , 1E- 10 , 7 , len);          int  index = 0 ;          double  minsa = Double.MAX_VALUE;          for  ( int  i = 0 ; i < km.repeat; i++) {              double  ss = km.run(i, datasource, len);              if  (ss < minsa) {                  minsa = ss;                  index = i;              }          }          System.out.println( "最好的结果是第"  + index + "次。" );      } }