k均值聚类算法

from numpy import *
import time
import matplotlib.pyplot as plt


def e_distance(v1,v2):#两点距离公式
    return sqrt(sum(power(v2-v1,2)))

def initcentroids(dataset,k):#聚类质心点初始化
    numsamples,dim = dataset.shape()
    centroids = zeros((k,dim))
    for i in range(k):
        index = int(random.uniform(0,numsamples))#随机选取样本中k个值作为当前聚类质心
        centroids[i,:]=dataset[index,:]
    return centroids

def kmeans(dataset,k):#K_mean算法
    numsamples = dataset.shape[0]
    clu_ass = mat(zeros((numsamples,2)))#第一列存储样本所属的簇，第二列存储对应距离
    clu_change = True
    centroids = initcentroids(dataset,k)#质心初始化
    while clu_change :
        clu_change = False
        for i in xrange(numsamples):
            min_dis = 1000000
            min_index = 0
            
            for j in range(k):
                distance = e_distance(centroids[j,:],dataset[i,:])
                if distance < min_dis:#更新最小距离以及对应的簇编号
                    min_dis = distance
                    min_index = j
            if clu_ass[i,0] != min_index:
                clu_change = True 
                clu_ass[i,:] = min_index,min_dis**2#更新ass列表
        for j in range(k):
            point_in_clu = dataset[nonzero(clu_ass[:,0].A == j)[0]]#属于j簇的所有样本这，句真难懂
            centroids[j,:] = mean(point_in_clu,axis = 0)#更新质心为所属簇样本的均值
    
                
    print 'mission complete'
    return centroids,clu_ass

暂时还没有运算结果和数据可视化的程序。关于数据可视化后面单独写一篇笔记好了..