层次聚类试图在不同层次上对数据集合进行划分, 从而形成树形的聚类结构。数据集的划分可采用“自底向上”的聚合策略,也可以采用“自顶向下”的分拆策略。
AGNES是一种采用自底向上的聚合策略的层次聚合算法,它先将数据集中的每个样本看作是一个初始的聚类簇,然后在算法进行的每一步中找出距离最近的两个聚类来进行合并,该过程不断的重复,直到到达预设的聚类簇的个数。
改算法的关键是如何计算聚类之间的距离, 实际上,每一个聚类是一个样本的集合,因此,只需要采用关于集合的某种距离即可。
最近距离由两个簇的最近的样本来决定, 最大距离由两个簇的最远的样本来决定,由此分别产生的AGNES算法又分别称为single-linkage和complete-linkage算法。
single-linkage算法
#include<stdio.h>
#include<stdlib.h>
#include<math.h>
#include<string.h>
#include<io.h>
int sample_size;
int sample_dimension;
char filename[200];
double** data;
double** distance;
int cluster_count;
void initialization();
void readDataFromFile();
double calculateDistance_BetweenTwoObject(int, int);
void initializationDistanceMatrix();
void AGNES();
void findMinDistance_BetweenClsuter_MIN(int*, int*);
void combineCluster(int, int);
void writeToFile(int);
int main(int argc, char* argv[])
{
if( argc != 4 )
{
printf("This algorithm requires 4 user-specified parameter"
"\n\t\tthe number of sample"
"\n\t\tthe dimension of sample"
"\n\t\tthe filename contain the sample"
"\n\n");
exit(0);
}
sample_size = atoi(argv[1]);
sample_dimension = atoi(argv[2]);
strcat(filename, argv[3]);
initialization();
readDataFromFile();
initializationDistanceMatrix();
AGNES();
return 0;
}
/*
* initialize the dynamic array for storing sample
* */
void initialization()
{
//initializaion the sample data
int i, j;
data = (double**)malloc(sizeof(double*) * (sample_size + 1));
if( !data )
{
printf("data malloc error: 0");
exit(0);
}
for( i = 1; i <= sample_size; i++ )
{
data[i] = (double*)malloc(sizeof(double) * (sample_dimension + 1));
if( !data[i] )
{
printf("data malloc error: %d", i);
exit(0);
}
}
//initialiation the distance data
distance = (double**)malloc(sizeof(double*) * (sample_size + 1));
if( !distance )
{
printf("distance malloc error: 0");
exit(0);
}
for( i = 1; i <= sample_size; i++ )
{
distance[i] = (double*)malloc(sizeof(double) * (sample_size + 1));
if( !distance[i] )
{
printf("distance malloc error: %d", i);
exit(0);
}
}
//the 0th element of all row indicate the clauterID of the object
for( i = 1; i <= sample_size; i++ )
{
distance[i][0] = i;
}
}
/*
* read the sample data from file
* */
void readDataFromFile()
{
FILE* fread;
int i;
int j;
if( NULL == (fread = fopen(filename, "r")))
{
printf("open file(%s) error: ", filename);
exit(0);
}
for( i = 1; i <= sample_size; i++ )
{
for( j = 1; j <= sample_dimension; j++ )
{
if( 1 != fscanf(fread, "%lf ", &data[i][j]))
{
printf("fscanf error: (%d, %d)", i, j);
exit(0);
}
}
}
//test
printf("print the origin data:\n");
for( i = 1; i <= sample_size; i++ )
{
for( j = 1; j <= sample_dimension; j++ )
{
printf("%f\t", data[i][j]);
}
printf("\n");
}
//test END
}
/*
* calculate distance between two objects
* */
double calculateDistance_BetweenTwoObject(int firstID, int secondID)
{
double distance = 0.0;
int i;
for( i = 1; i <= sample_dimension; i++ )
{
distance = distance + pow(data[firstID][i] - data[secondID][i], 2);
}
return sqrt(distance);
}
/*
* calculate initialization distance matrix
* */
void initializationDistanceMatrix()
{
int i, j;
for( i = 1; i <= sample_size; i++ )
{
for( j = i; j <= sample_size; j++ )
{
distance[i][j] = calculateDistance_BetweenTwoObject(i, j);
distance[j][i] = distance[i][j];
}
}
//test
printf("print the origin distance matrix\n");
for( i = 1; i <= sample_size; i++ )
{
for( j = 0; j <= sample_size; j++ )
{
printf("%f ", distance[i][j]);
}
printf("\n");
}
//test END
}
/***************************************************************************************************************
* AGNES
***************************************************************************************************************/
void AGNES()
{
cluster_count = sample_size;
int cluster_1, cluster_2;
int count = 1;
while( cluster_count > 3 )
{
printf("-------------------%d--------------------\n", count);
findMinDistance_BetweenClsuter_MIN(&cluster_1, &cluster_2);
combineCluster(cluster_1, cluster_2);
cluster_count--;
writeToFile(count++);
}
}
/*
* find two clusters the distance between them is minimun, and store the clusterID in @cluster_1 and @cluster_2, respectively.
* */
void findMinDistance_BetweenClsuter_MIN(int* cluster_1, int* cluster_2)
{
int i, j;
double min_distance;
int flag = 1;
for( i = 1; i <= sample_size; i++ )
{
for( j = i + 1; j <= sample_size; j++ )
{
if( distance[i][0] == distance[j][0] )
{
printf("same cluster!!!! %d and %d\n\n", i , j);
continue;
}
else if( flag == 1 )
{
min_distance = distance[i][j];
*cluster_1 = i;
*cluster_2 = j;
flag = 0;
printf("----get the initial minimum distance is %f(%d, %d)\n", min_distance, i, j);
continue;
}
if( distance[i][j] < min_distance )
{
min_distance = distance[i][j];
*cluster_1 = i;
*cluster_2 = j;
}
}
}
printf("the minimum is %f :(%d, %d)\n", min_distance, *cluster_1, *cluster_2);
}
/*
* combine the two clusters to one
* */
void combineCluster(int cluster_1, int cluster_2)
{
int i;
int buffer;
buffer = distance[cluster_2][0];
for( i = 1; i <= sample_size; i++ )
{
if( distance[i][0] == buffer )
{
distance[i][0] = distance[cluster_1][0];
}
}
//test
printf(" the minimum distance is %f, and the object is %d and %d\n", distance[cluster_1][cluster_2], cluster_1, cluster_2);
int j;
for( i = 1; i <= sample_size; i++ )
{
for( j = 0; j <= sample_size; j++ )
{
printf("%f ", distance[i][j]);
}
printf("\n");
}
}
/*
* write the result of clustering information to file
* */
void writeToFile(int round)
{
int i;
int j;
int* auxiliary = (int*)malloc(sizeof(int) * (sample_size + 1));
if( !auxiliary )
{
printf("auxiliary malloc error");
exit(0);
}
for( i = 1; i <= sample_size; i++ )
auxiliary[i] = 0;
for( i = 1; i <= sample_size; i++ )
{
auxiliary[(int)distance[i][0]]++;
}
int *clusterID = (int*)malloc(sizeof(int) * (sample_size - round + 1));
if( !clusterID )
{
printf("clusterID malloc error");
exit(0);
}
int counter = 1;
for( i = 1; i <= sample_size; i++ )
{
if( auxiliary[i] != 0 )
clusterID[counter++] = i;
}
//test
printf("cluster ID:");
for( i = 1; i <= sample_size - round; i++ )
{
printf("%d ", clusterID[i]);
}
printf("\n");
//test END
printf("writeToFile -----------round: %d\n", round);
FILE** fwrite;
fwrite = (FILE**)malloc(sizeof(FILE*) * (sample_size + 1));
if( !fwrite )
{
printf("fwrite malloc error\n");
exit(0);
}
char filename[200] = "";
char instruction[200] = "";
sprintf(instruction, "md Round_%d", round);
system(instruction);
for( i = 1; i <= sample_size - round; i++ )
{
sprintf(filename, ".//Round_%d//cluster_%d.data", round, clusterID[i]);
if( NULL == (fwrite[clusterID[i]] = fopen(filename, "w")))
{
printf("open file(%s) error\n", filename);
exit(0);
}
}
for( i = 1; i <= sample_size; i++ )
{
printf("%d ", (int)distance[i][0]);
for( j = 1; j <= sample_dimension; j++ )
{
fprintf(fwrite[(int)distance[i][0]], "%f\t", data[i][j]);
printf("%f ", data[i][j]);
}
fprintf(fwrite[(int)distance[i][0]], "\n");
printf("\n");
}
for( i = 1; i <= sample_size - round; i++ )
{
fclose(fwrite[clusterID[i]]);
}
}
complete-linkage算法
#include<stdio.h>
#include<stdlib.h>
#include<math.h>
#include<string.h>
#include<io.h>
int sample_size;
int sample_dimension;
char filename[200];
double** data;
double** distance;
int cluster_count;
void initialization();
void readDataFromFile();
double calculateDistance_BetweenTwoObject(int, int);
void initializationDistanceMatrix();
void findMinDistance_BetweenClsuter_MAX(int*, int*);
void AGNES();
void combineCluster(int, int);
void updateDistance(int);
void writeToFile(int);
int main(int argc, char* argv[])
{
if( argc != 4 )
{
printf("This algorithm requires 4 user-specified parameter"
"\n\t\tthe number of sample"
"\n\t\tthe dimension of sample"
"\n\t\tthe filename contain the sample"
"\n\n");
exit(0);
}
sample_size = atoi(argv[1]);
sample_dimension = atoi(argv[2]);
strcat(filename, argv[3]);
initialization();
readDataFromFile();
initializationDistanceMatrix();
AGNES();
return 0;
}
/*
* initialize the dynamic array for storing sample
* */
void initialization()
{
//initializaion the sample data
int i, j;
data = (double**)malloc(sizeof(double*) * (sample_size + 1));
if( !data )
{
printf("data malloc error: 0");
exit(0);
}
for( i = 1; i <= sample_size; i++ )
{
data[i] = (double*)malloc(sizeof(double) * (sample_dimension + 1));
if( !data[i] )
{
printf("data malloc error: %d", i);
exit(0);
}
}
//initialiation the distance data
distance = (double**)malloc(sizeof(double*) * (sample_size + 1));
if( !distance )
{
printf("distance malloc error: 0");
exit(0);
}
for( i = 0; i <= sample_size; i++ )
{
distance[i] = (double*)malloc(sizeof(double) * (sample_size + 1));
if( !distance[i] )
{
printf("distance malloc error: %d", i);
exit(0);
}
}
//the 0th element of all row indicate the clauterID of the object
for( i = 1; i <= sample_size; i++ )
{
distance[i][0] = i;
}
//where reserved if distance[0][i] = 1,
for( i = 1; i <= sample_size; i++ )
{
distance[0][i] = 1;
}
}
/*
* read the sample data from file
* */
void readDataFromFile()
{
FILE* fread;
int i;
int j;
if( NULL == (fread = fopen(filename, "r")))
{
printf("open file(%s) error: ", filename);
exit(0);
}
for( i = 1; i <= sample_size; i++ )
{
for( j = 1; j <= sample_dimension; j++ )
{
if( 1 != fscanf(fread, "%lf ", &data[i][j]))
{
printf("fscanf error: (%d, %d)", i, j);
exit(0);
}
}
}
}
/*
* calculate distance between two objects
* */
double calculateDistance_BetweenTwoObject(int firstID, int secondID)
{
double distance = 0.0;
int i;
for( i = 1; i <= sample_dimension; i++ )
{
distance = distance + pow(data[firstID][i] - data[secondID][i], 2);
}
return sqrt(distance);
}
/*
* calculate initialization distance matrix
* */
void initializationDistanceMatrix()
{
int i, j;
for( i = 1; i <= sample_size; i++ )
{
for( j = i; j <= sample_size; j++ )
{
distance[i][j] = calculateDistance_BetweenTwoObject(i, j);
distance[j][i] = distance[i][j];
}
}
}
/***************************************************************************************************************
* AGNES
***************************************************************************************************************/
void AGNES()
{
cluster_count = sample_size;
int cluster_1, cluster_2;
int count = 1;
while( cluster_count > 3 )
{
printf("-------------------%d--------------------\n", count);
findMinDistance_BetweenClsuter_MAX(&cluster_1, &cluster_2);
combineCluster(cluster_1, cluster_2);
updateDistance(cluster_1);
cluster_count--;
writeToFile(count++);
}
}
/*
* find two clusters the distance between them is minimun, and store the clusterID in @cluster_1 and @cluster_2, respectively.
* */
void findMinDistance_BetweenClsuter_MAX(int* cluster_1, int* cluster_2)
{
int i, j;
double min_distance;
int flag = 1;
int object_1;
int object_2;
for( i = 1; i <= sample_size; i++ )
{
for( j = i + 1; j <= sample_size; j++ )
{
if( distance[i][0] == distance[j][0] || ( distance[0][i] == 0 || distance[0][j] == 0 ) )
{
continue;
}
else if( flag == 1 )
{
min_distance = distance[i][j];
object_1 = i;
object_2 = j;
flag = 0;
continue;
}
if( distance[i][j] < min_distance )
{
min_distance = distance[i][j];
object_1 = i;
object_2 = j;
}
}
}
*cluster_1 = distance[object_1][0];
*cluster_2 = distance[object_2][0];
}
/*
* combine the two clusters to one
* */
void combineCluster(int cluster_1, int cluster_2)
{
int i;
for( i = 1; i <= sample_size; i++ )
{
if( distance[i][0] == cluster_2 )
{
distance[i][0] = cluster_1;
}
}
}
/*
* update distance data
* */
void updateDistance(int clusterID)
{
int i, j;
int flag = 1;
int save_i;
for( i = 1; i <= sample_size; i++ )
{
if( distance[i][0] == clusterID && flag == 1 )
{
flag = 0;
save_i = i;
continue;
}
else if( distance[i][0] == clusterID && flag == 0 )
{
distance[0][i] = 0;
for( j = 1; j <= sample_size; j++ )
{
if( distance[i][j] > distance[save_i][j] )
{
distance[save_i][j] = distance[i][j];
}
}
}
}
for( i = 1; i <= sample_size; i++ )
{
distance[i][save_i] = distance[save_i][i];
}
}
/*
* write the result of clustering information to file
* */
void writeToFile(int round)
{
int i;
int j;
int* auxiliary = (int*)malloc(sizeof(int) * (sample_size + 1));
if( !auxiliary )
{
printf("auxiliary malloc error");
exit(0);
}
for( i = 1; i <= sample_size; i++ )
auxiliary[i] = 0;
for( i = 1; i <= sample_size; i++ )
{
auxiliary[(int)distance[i][0]]++;
}
int *clusterID = (int*)malloc(sizeof(int) * (sample_size - round + 1));
if( !clusterID )
{
printf("clusterID malloc error");
exit(0);
}
int counter = 1;
for( i = 1; i <= sample_size; i++ )
{
if( auxiliary[i] != 0 )
clusterID[counter++] = i;
}
printf("writeToFile -----------round: %d\n", round);
FILE** fwrite;
fwrite = (FILE**)malloc(sizeof(FILE*) * (sample_size + 1));
if( !fwrite )
{
printf("fwrite malloc error\n");
exit(0);
}
char filename[200] = "";
char instruction[200] = "";
sprintf(instruction, "md Round_%d", round);
system(instruction);
for( i = 1; i <= sample_size - round; i++ )
{
sprintf(filename, ".//Round_%d//cluster_%d.data", round, clusterID[i]);
if( NULL == (fwrite[clusterID[i]] = fopen(filename, "w")))
{
printf("open file(%s) error\n", filename);
exit(0);
}
}
for( i = 1; i <= sample_size; i++ )
{
printf("%d ", (int)distance[i][0]);
for( j = 1; j <= sample_dimension; j++ )
{
fprintf(fwrite[(int)distance[i][0]], "%f\t", data[i][j]);
}
fprintf(fwrite[(int)distance[i][0]], "\n");
}
for( i = 1; i <= sample_size - round; i++ )
{
fclose(fwrite[clusterID[i]]);
}
}