机器学习中K-means聚类算法原理及C语言实现

本人以前主要focus在传统音频的软件开发,接触到的算法主要是音频信号处理相关的,如各种编解码算法和回声消除算法等。最近切到语音识别上,接触到的算法就变成了各种机器学习算法,如GMM等。K-means作为其中比较简单的一种肯定是要好好掌握的。今天就讲讲K-means的基本原理和代码实现。其中基本原理简述(主要是因为:1,K-means比较简单;2,网上有很多讲K-means基本原理的),重点放在代码实现上。

 

1, K-means基本原理

K均值(K-means)聚类算法是无监督聚类(聚类(clustering)是将数据集中的样本划分为若干个通常是不相交的子集,每个子集称为一个“簇(cluster)”)算法中的一种,也是最常用的聚类算法。K表示类别数,Means表示均值。K-means主要思想是在给定K值和若干样本(点)的情况下,把每个样本(点)分到离其最近的类簇中心点所代表的类簇中,所有点分配完毕之后,根据一个类簇内的所有点重新计算该类簇的中心点(取平均值),然后再迭代的进行分配点和更新类簇中心点的步骤,直至类簇中心点的变化很小,或者达到指定的迭代次数。

 

 K-means算法流程如下:

(a)随机选取K个初始cluster center

(b)分别计算所有样本到这K个cluster center的距离

(c)如果样本离cluster center Ci最近,那么这个样本属于Ci点簇;如果到多个cluster center的距离相等,则可划分到任意簇中

(d)按距离对所有样本分完簇之后,计算每个簇的均值(最简单的方法就是求样本每个维度的平均值),作为新的cluster center

(e)重复(b)(c)(d)直到新的cluster center和上轮cluster center变化很小或者达到指定的迭代次数,算法结束

 

2, 算法实现

我主要偏底层开发,最熟悉语言是C,所以代码是用C语言来实现的。在二维平面上有一些点,大意如下图,

机器学习中K-means聚类算法原理及C语言实现_第1张图片

用K-means算法对其分类,其中类的个数(即K值)和点的个数人为指定。具体的代码如下:

#include
#include
#include
#include

#define MAX_ROUNDS 100    //最大允许的聚类次数

//“点”的结构体  
typedef struct Point{
  int x_value;           //用于存放点在X轴上的值
  int y_value;           //用于存放点在Y轴上的值
  int cluster_id;        //用于存放该点所属的cluster id
}Point;
Point* data;
 
//cluster center的结构体
typedef struct ClusterCenter{
  double x_value;
  double y_value;
  int cluster_id;
}ClusterCenter;
ClusterCenter* cluster_center;

//计算cluster center的结构体
typedef struct CenterCalc{
  double x_value;
  double y_value;
}CenterCalc;
CenterCalc *center_calc;
 
int is_continue;                               //kmeans 运算是否继续
int* cluster_center_init_index;        //记录每个cluster center最初用的是哪个“点”
double* distance_from_center;      //记录一个“点”到所有cluster center的距离
int* data_size_per_cluster;            //每个cluster点的个数
int data_size_total;                        //设定点的个数
char filename[200];                       //要读取的点的数据的文件名
int cluster_count;                          //设定的cluster的个数
 
void memoryAlloc();
void memoryFree();
void readDataFromFile();
void initialCluster();
void calcDistance2OneCenter(int pointID, int centerID);
void calcDistance2AllCenters(int pointID);
void partition4OnePoint(int pointID);
void partition4AllPointOneCluster();
void calcClusterCenter();
void kmeans();
void compareNewOldClusterCenter(CenterCalc* center_calc);
 
int main(int argc, char* argv[])
{
    if( argc != 4 )
    {
        printf("This application needs 3 parameters to run:"
            "\n the 1st is the size of data set,"
            "\n the 2nd is the file name that contains data"
            "\n the 3rd indicates the cluster_count"
            "\n");
        exit(1);
    }

    data_size_total = atoi(argv[1]);
    strcat(filename, argv[2]);
    cluster_count = atoi(argv[3]);
    //1, memory alloc
    memoryAlloc();
    //2, read point data from file
    readDataFromFile();
    //3, initial cluster
    initialCluster();
    //4, run k-means
    kmeans();
    //5, memory free & end
    memoryFree();
    
    return 0;
}

void memoryAlloc()
{
  data = (Point*)malloc(sizeof(struct Point) * (data_size_total));
  if( !data )
  {
    printf("malloc error:data!");
    exit(1);
  }
  cluster_center_init_index = (int*)malloc(sizeof(int) * (cluster_count));
  if( !cluster_center_init_index )
  {
    printf("malloc error:cluster_center!\n");
    exit(1);
  }
  distance_from_center = (double*)malloc(sizeof(double) * (cluster_count));
  if( !distance_from_center )
  {
    printf("malloc error: distance_from_center!\n");
    exit(1);
  }
  cluster_center = (ClusterCenter*)malloc(sizeof(struct ClusterCenter) * (cluster_count));
  if( !cluster_center )
  {
    printf("malloc cluster center new error!\n");
    exit(1);
  }

  center_calc = (CenterCalc*)malloc(sizeof(CenterCalc) * cluster_count);
  if( !center_calc )
  {
    printf("malloc error: center_calc!\n");
    exit(1);
  }

  data_size_per_cluster = (int*)malloc(sizeof(int) * (cluster_count));
  if( !data_size_per_cluster )
  {
    printf("malloc error: data_size_per_cluster\n");
    exit(1);
  }
 
}

void memoryFree()
{
  free(data);
  data = NULL;
  free(cluster_center_init_index);
  cluster_center_init_index = NULL;
  free(distance_from_center);
  distance_from_center = NULL;
  free(cluster_center);
  cluster_center = NULL;
  free(center_calc);
  center_calc = NULL;
  free(data_size_per_cluster);
  data_size_per_cluster = NULL;
}

//从文件中读入每个点的x和y值
void readDataFromFile()
{
  int i;
  FILE* fread;
 
  if( NULL == (fread = fopen(filename, "r")))
  {
    printf("open file(%s) error!\n", filename);
    exit(1);
  }

  for( i = 0; i < data_size_total; i++ )
  {
    if( 2 != fscanf(fread, "%d %d ", &data[i].x_value, &data[i].y_value))
    {
      printf("fscanf error: %d\n", i);
    }
    data[i].cluster_id = -1;    //初始时每个点所属的cluster id均置为-1

    printf("After reading, point index:%d, X:%d, Y:%d, cluster_id:%d\n", i, data[i].x_value, data[i].y_value, data[i].cluster_id);
  }
}
 

//根据传入的cluster_count来随机的选择一个点作为 一个cluster的center  
void initialCluster()
{
  int i,j;
  int random;
    
  //产生初始化的cluster_count个聚类  
  for( i = 0; i < cluster_count; i++ )
  {
    cluster_center_init_index[i] = -1;
  }
  //随机选择一个点作为每个cluster的center(不重复)
  for( i = 0; i < cluster_count; i++ )
  {
    Reselect:
        random = rand() % (data_size_total - 1);
        for(j = 0; j < i; j++) {
            if(random == cluster_center_init_index[j])
                goto Reselect;
        }

    cluster_center_init_index[i] = random;
    printf("cluster_id: %d, located in point index:%d\n", i, random);  
  }
  //将随机选择的点作为center,同时这个点的cluster id也就确定了
  for( i = 0; i < cluster_count; i++ )
  {
    cluster_center[i].x_value = data[cluster_center_init_index[i]].x_value;
    cluster_center[i].y_value = data[cluster_center_init_index[i]].y_value;
    cluster_center[i].cluster_id = i;
    data[cluster_center_init_index[i]].cluster_id = i;

    printf("cluster_id:%d, index:%d, x_value:%f, y_value:%f\n", cluster_center[i].cluster_id, cluster_center_init_index[i], cluster_center[i].x_value, cluster_center[i].y_value);
  }
}
 

//计算一个点到一个cluster center的distance
void calcDistance2OneCenter(int point_id,int center_id)
{
  distance_from_center[center_id] = sqrt( (data[point_id].x_value-cluster_center[center_id].x_value)*(double)(data[point_id].x_value-cluster_center[center_id].x_value) + (double)(data[point_id].y_value-cluster_center[center_id].y_value) *              (data[point_id].y_value-cluster_center[center_id].y_value) );
}
 
//计算一个点到每个cluster center的distance
void calcDistance2AllCenters(int point_id)
{
  int i;
  for( i = 0; i < cluster_count; i++ )
  {
    calcDistance2OneCenter(point_id, i);
  }
}
 
//确定一个点属于哪一个cluster center(取距离最小的)
void partition4OnePoint(int point_id)
{
  int i;
  int min_index = 0;
  double min_value = distance_from_center[0];
  for( i = 0; i < cluster_count; i++ )
  {
    if( distance_from_center[i] < min_value )
    {
      min_value = distance_from_center[i];
      min_index = i;
    }
  }
 
  data[point_id].cluster_id = cluster_center[min_index].cluster_id;
}

//在一轮的聚类中得到所有的point所属于的cluster center
void partition4AllPointOneCluster()
{
  int i;
  for( i = 0; i < data_size_total; i++ )
  {
    if( data[i].cluster_id != -1 )  //这个点就是center,不需要计算
      continue;
    else
    {
      calcDistance2AllCenters(i);  //计算第i个点到所有center的distance
      partition4OnePoint(i);          //根据distance对第i个点进行partition
    }
  }
}

//重新计算新的cluster center
void calcClusterCenter()
{
  int i;

  memset(center_calc, 0, sizeof(CenterCalc) * cluster_count);
  memset(data_size_per_cluster, 0, sizeof(int) * cluster_count);
  //分别对每个cluster内的每个点的X和Y求和,并计每个cluster内点的个数
  for( i = 0; i < data_size_total; i++ )
  {
    center_calc[data[i].cluster_id].x_value += data[i].x_value;
    center_calc[data[i].cluster_id].y_value += data[i].y_value;
    data_size_per_cluster[data[i].cluster_id]++;
  }
  //计算每个cluster内点的X和Y的均值作为center
  for( i = 0; i < cluster_count; i++ )
  {
     if(data_size_per_cluster[i] != 0) {
        center_calc[i].x_value = center_calc[i].x_value/ (double)(data_size_per_cluster[i]);
        center_calc[i].y_value = center_calc[i].y_value/ (double)(data_size_per_cluster[i]);

        printf(" cluster %d point cnt:%d\n", i, data_size_per_cluster[i]);
        printf(" cluster %d center: X:%f, Y:%f\n", i, center_calc[i].x_value, center_calc[i].y_value);
    }
    else
          printf(" cluster %d count is zero\n", i);
  }
 
  //比较新的和旧的cluster center值的差别。如果是相等的,则停止K-means算法。
  compareNewOldClusterCenter(center_calc);
 
  //将新的cluster center的值放入cluster_center结构体中
  for( i = 0; i < cluster_count; i++ )
  {
    cluster_center[i].x_value = center_calc[i].x_value;
    cluster_center[i].y_value = center_calc[i].y_value;
    cluster_center[i].cluster_id = i;
  }

  //在重新计算了新的cluster center之后,要重新来为每一个Point进行聚类,所以data中用于表示聚类ID的cluster_id要都重新置为-1。
  for( i = 0; i < data_size_total; i++ )
  {
    data[i].cluster_id = -1;
  }
}
 
//比较新旧的cluster center的值,完全一样表示聚类完成
void compareNewOldClusterCenter(CenterCalc* center_calc)
{
  int i;
  is_continue = 0;       //等于0表示不要继续,1表示要继续
  for( i = 0; i < cluster_count; i++ )
  {
    if( center_calc[i].x_value != cluster_center[i].x_value || center_calc[i].y_value != cluster_center[i].y_value)
    {
      is_continue = 1;
      break;
    }
  }
}
 
//K-means算法
void kmeans()
{
  int rounds;
  for( rounds = 0; rounds < MAX_ROUNDS; rounds++ )
  {
    printf("\nRounds : %d             \n", rounds+1);
    partition4AllPointOneCluster();
    calcClusterCenter();
    if( 0 == is_continue )
    {
       printf("\n after %d rounds, the classification is ok and can stop.\n", rounds+1);
       break;  
    }
  }
}
 

编译后生成可执行文件kmeans,输入的文件里共有6个点,分别为(0, 0), (4, 4), (4, 5), (0, 1), (3, 6) ,(4, 9),要求分成两类。运行可执行程序后得到结果如下:

$ ./kmeans 6 data 2
After reading, point index:0, X:0, Y:0, cluster_id:-1
After reading, point index:1, X:4, Y:4, cluster_id:-1
After reading, point index:2, X:4, Y:5, cluster_id:-1
After reading, point index:3, X:0, Y:1, cluster_id:-1
After reading, point index:4, X:3, Y:6, cluster_id:-1
After reading, point index:5, X:4, Y:9, cluster_id:-1


cluster_id: 0, located in point index:3
cluster_id: 1, located in point index:1
cluster_id:0, index:3, x_value:0.000000, y_value:1.000000
cluster_id:1, index:1, x_value:4.000000, y_value:4.000000

Rounds : 1             
 cluster 0 point cnt:2
 cluster 0 center: X:0.000000, Y:0.500000
 cluster 1 point cnt:4
 cluster 1 center: X:3.750000, Y:6.000000

Rounds : 2             
 cluster 0 point cnt:2
 cluster 0 center: X:0.000000, Y:0.500000
 cluster 1 point cnt:4
 cluster 1 center: X:3.750000, Y:6.000000

 after 2 rounds, the classification is ok and can stop.


即两轮后聚类就好了,(0, 0),(0, 1)一类,(4, 4), (4, 5), (3, 6) ,(4, 9)一类。

你可能感兴趣的:(智能语音,K-means)