I also tested it using python, and the clustering was completed in one round. It may also be because the data set is incomplete. There are only dozens of data sets, so there is only one round.
(Update):
This can be combined with the random selection of centroids in the algorithm. It is related to the algorithm, because different centroids have an impact on the final number of iterations, just like the subject said to achieve the local optimal solution
The following can be cycled three times

Cycle six times:

Core part of the algorithm:

def k_means(data_set, k, distE=distEclud, createCent=rand_cent): """ data_set: 数据集 data_set = [[1.658985, 4.285136], [-3.453687, 3.424321], [4.838138, -1.151539],[-5.379713, -3.362104], [0.972564, 2.924086]] k: 设置聚类簇数K值 distE：函数, 计算数据点的欧式距离 例如：vector1 = [[ 1.658985 4.285136]] vector2 = [[-3.453687 3.424321]] 欧氏距离：distance = distE(vector1, vector2) createCent: 函数， 返回一个随机的质点矩阵 """ m = shape(data_set)[0] # 获取行数 clusterAssment = mat(zeros((m, 2))) # 初始化一个矩阵， 用来记录簇索引和存储误差平方和(指当前点到簇质点的距离） centroids = rand_cent(data_set, k) # 随机生成一个质心矩阵蔟 clusterChanged = True print('开始') while clusterChanged: clusterChanged = False for i in range(m): # 对每个数据点寻找最近的质心 min_dist = inf # 设置最小距离为正无穷大 min_index = -1 for j in range(k): # 遍历质心簇，寻找最近质心 dist_J = distEclud(centroids[j, :], data_set[i, :]) if dist_J < min_dist: min_dist = dist_J min_index = j if clusterAssment[i, 0] != min_index: clusterChanged = True clusterAssment[i, :] = min_index, min_dist ** 2 # 平方的意义在于判断聚类结果的好坏 print(centroids) print('==================') for cent in range(k): # 更新质心，将每个簇中的点的均值作为质心 index_all = clusterAssment[:, 0].A # 取出样本所属簇的索引值 value = nonzero(index_all == cent) # 取出所有属于第cent个簇的索引值 sampleInClust = data_set[value[0]] # 取出属于第I个簇的所有样本点 centroids[cent, :] = mean(sampleInClust, axis=0) return centroids, clusterAssment

Because the selection of initial points is random, of course the clustering process may not be exactly the same