def printBasicInfo(centroids,clusterAssment,norMat): minCount = 10000 maxCount = 0 cluster_label = clusterAssment[:,0] clusters = [[],[],[],[]] for i in range(0,len(cluster_label)): clusters[int(cluster_label[i])].append(np.asarray(norMat)[i]) clusters = np.asarray(clusters) max_cluster = 0 min_cluster = 0 for i in range(0,len(clusters)): if minCount > len(clusters[i]): minCount = len(clusters[i]) if maxCount < len(clusters[i]): maxCount = len(clusters[i]) if len(clusters[max_cluster]) < len(clusters[i]): max_cluster = i if len(clusters[min_cluster]) > len(clusters[i]): min_cluster = i print "%d cluster has %d elements " % (i, len(clusters[i])), print "the centroids is", print centroids[i] number_weight = float(len(clusters[max_cluster]))/(len(clusters[min_cluster])) print centroids[max_cluster] di = base.dunn(clusters) NDunnIndex = di*(maxCount/minCount) print "original dunn is %f" % di print "weighted dunn is %f" % (number_weight*di)
def main():
dataSet = kMeans.loadfromcsv('./data/8.csv')
dataMat = np.mat(dataSet)
# normalize dataMat
norMat = kMeans.normalize(dataMat)
# centroids is the center of clusters
# clusterAssment[cluster_index,deviation],in which deviation represents the dist
# from current point to centroids.
# 使用Bisceting Kmeans算法对游客进行聚类，预期聚类簇数目为4
centroids, clusterAssment = kMeans.biKmeans(norMat,4)
cluster_label = clusterAssment[:,0]
clusters = [[],[],[],[]]
for i in range(0,len(cluster_label)):
clusters[(int)(cluster_label[i])].append(np.asarray(norMat)[i])
clusters = np.asarray(clusters)
for i in range(0,len(clusters)):
clusters[i] = np.asarray(clusters[i])
clusters = np.asarray(clusters)
#找出含有元素最多的簇，以及最少的簇
minCount = 10000
maxCount = 0
max_cluster = 0
min_cluster = 0
for i in range(0,len(clusters)):
if minCount > len(clusters[i]):
minCount = len(clusters[i])
if maxCount < len(clusters[i]):
maxCount = len(clusters[i])
if len(clusters[max_cluster]) < len(clusters[i]):
max_cluster = i
if len(clusters[min_cluster]) > len(clusters[i]):
min_cluster = i
print "%d cluster has %d elements " % (i, len(clusters[i])),
print "the centroids is",
print centroids[i]
number_weight = float(len(clusters[max_cluster]))/(len(clusters[min_cluster]))
print centroids[max_cluster]
# 计算Dunn指标
di = base.dunn(clusters)
# 计算N-Dunn指标
NDunnIndex = di*(maxCount/minCount)
print di
print NDunnIndex
print "original dunn is %f" % di
print "weighted dunn is %f" % (number_weight*di)
