关于#python#的问题，如何解决？

Question

python在跑k均值聚类时候的时候，每次运行结果都不一样，求帮忙瞅瞅是哪出问题了？

import numpy as np
from sklearn.datasets import load_iris
iris = load_iris()
x,y =iris.data,iris.target
#x=np.array([[1,1,1,1],[10,10,10,10],[20,20,20,20],[5,5,5,5],[3,3,3,3],[6,6,6,6],[25,25,25,25]])
centroids_a=[]
list1=[]
#聚类中心
def init_random_centroids(k,x):
    data1=x.shape[0]
    suiji_a =np.random.choice(data1,k,replace=False)
    for i in suiji_a:
        xlist = x[i]
        centroids_a.append(xlist)
        centroids =np.array(centroids_a)
    return centroids


#计算距离
def euclidean_distance(one_sample, x):
    #print("中心点",x)
    #print("样本",one_sample)
    #print("单个样本",one_sample)
    #print("中心",x)
    distances= np.sum((x-one_sample)**2)
    #print("距离",distances)
    return distances


#返回离样本最近的中心索引
def _closest_centroid(sample, centroids):
    k = centroids.shape[0]
    #print("k",k)
    #print("聚类中心",centroids)
    list_distance =[]
    for i in range(k):
        a=euclidean_distance(one_sample=sample, x=centroids[i])
        list_distance.append(a)
        # print("list",list_distance)
        distance_list=np.array(list_distance)
        id =np.argmin(distance_list)
    #print("单个样本与聚类中心的距离：",distance_list)
    #print("距离最近的中心索引",id)
    return id

#将所有样本进行归类，归类规则就是将该样本归类到与其最近的中心
def create_clusters(k,centroids, x):
    clusters = [[] for _ in range(k)]
    #print("cen",centroids)
    for sample in x:
        ys=_closest_centroid(sample, centroids)
        #print(ys)
        clusters[ys].append(sample)
    #print("0",clusters[0])
    #print("1",clusters[1])
    #print("2",clusters[2])
    return clusters

#中心点更新
def update_centroids(k,clusters):
    #print(k)
    # print("类型0",clusters[0])
    # print("类型1",clusters[1])
    # print("类型2",clusters[2])
    for i in range(k):
        centroid = np.mean(clusters[i], axis=0)
        #print("平均聚点",centroid)
        centroids[i] = centroid
    print("新聚类中心",centroids)
    return centroids


# 将所有样本进行归类，其所在的类别的索引就是其类别标签
def get_cluster_labels(clusters, x):
    #print(clusters)
    y_pred = []
    for sample in x:
        ys=_closest_centroid(sample=sample, centroids=centroids)
        y_pred.append(ys)
    print(y_pred)
    return y_pred



#随机选取k个聚类中心
centroids = init_random_centroids(3, x)
for number in range(20000000):
    #样本归类
    cluster = create_clusters(k=3,centroids=centroids,x=x)
    former_centroids = centroids
    #更新新的聚类中心
    clusters = update_centroids(k=3,clusters=cluster)
    diff = centroids - former_centroids
    if diff.any() <0.000001:
        get_cluster_labels(clusters=cluster,x=x)
        break

Answer 1

K-Means算法从随机初始化簇质心开始。每次运行KMeans时，此选择都会有所不同，可能会产生不同的结果。为了得到可再现的结果，可以在KMeans中使用random_state参数，这将修复簇中心线的初始选择：

model = KMeans(n_clusters=number_of_clusters, 
               init='k-means++', 
               max_iter=100, 
               n_init=100, 
               random_state=123)