python利用K均值做聚类,判断中国足球第几流

读了博客园的一篇 文章

受到启发,写了一个K均值的python实现,代码如下:

import random
from math import sqrt

sample=[
    [1,1,0.5],
    [0.3,0,0.19],
    [0,0.15,0.13],
    [0.24,0.76,0.25],
    [0.3,0.76,0.06],
    [1,1,0],
    [1,0.76,0.5],
    [1,0.76,0.5],
    [0.7,0.76,0.25],
    [1,1,0.5],
    [1,1,0.25],
    [1,1,0.5],
    [0.7,0.76,0.5],
    [0.7,0.68,0.5],
    [1,1,0.5]
]

samplename=['中国','日本','韩国','伊朗','沙特','伊拉克','卡塔尔','阿联酋','乌兹别克斯坦','泰国','越南','阿曼','巴林','朝鲜','印尼']

def EDistance(v1,v2):
    tmp=sum([pow(v1[i]-v2[i],2) for i in range(len(v1))])
    return sqrt(tmp)


class kcluster:
    k=3
    distance=mypearson
    rows=sample

    #获取用于比较的序列的在各个维度上均值组成的序列
    def getavg(self,rows,seed):
        n=len(rows)
        if n==0:
            return seed
        rs=[]
        for i in range(len(rows[0])):
            rs.append(sum([row[i] for row in rows])/n)
        return rs

    #根据种子获取与种子最接近的序列
    def getbestmatch(self,rows,seeds):
        bestmatch={}
        for i in range(self.k):
            bestmatch.setdefault(i,[])
        #判断每个序列最匹配的种子
        for row in rows:
            d=9999
            whichseed=0
            i=0
            for seed in seeds:
                tmp=EDistance(row,seed)
                if tmp<d:
                    d=tmp
                    whichseed=i
                i+=1
            bestmatch[whichseed].append(row)
        return bestmatch

    #生成随机种子
    def getseeds(self):
        #每个维度上最值组成的元组
        minandmax=[]
        for i in range(len(self.rows[0])):
            minandmax.append((min([row[i] for row in self.rows]),max([row[i] for row in self.rows])))
        seeds=[]
        for i in range(self.k):
            #生成随机种子
            seeds.append([random.random()*(row[1]-row[0])+row[0] for row in minandmax])
        return seeds

    #K均值聚类的主函数
    def kcluster(self):
        #生成种子
        seeds=self.getseeds()
        lastseeds=seeds[:]
        while True:
            #根据种子生成最佳聚类
            bestmatch=self.getbestmatch(self.rows,seeds)
            #print(seeds)
            #print(bestmatch)

            #移动种子到匹配序列的均值处
            for i in range(self.k):
               seeds[i]=self.getavg(bestmatch[i],seeds[i])
            #print(seeds)
            #print(lastseeds)

            if lastseeds==seeds:
                break
            else: lastseeds=seeds[:]
        return bestmatch


obj=kcluster()
rs=obj.kcluster()
print(rs)

for j in range(obj.k):
    for i in range(len(sample)):
        if sample[i] in rs[j]:
            print(samplename[i],end=' ')
    print('')

运行后你会发现两点:

1. 种子的选择会对聚类结果造成很大的影响

2. 但无论种子怎么选,中国足球都是三流