聚类

import pandas as pd 
import numpy as np
import matplotlib.pyplot as plt
def LoadData(filename):
    data=pd.read_csv(filename)#dataframe类型
    return data.values[:,0:4]#二维数组， n1:n2 取从n1开始，不包括n2的行或列。只取值，不取标签
#初始化k个centroid
def randCent(dataSet, k):#分k类
    n = np.shape(dataSet)[1]#它的功能是读取矩阵的长度,二维的返回（行数,列数）取列数
    centroids = np.mat(np.zeros((k,n)))   # 每个质心有n个坐标值，总共要k个质心 zeros的第一个参数是(k,n) k行n列（3*4）
    for j in range(n):#初始化n个维度。每次是一列一起操作的
        minJ = min(dataSet[:,j])
        maxJ = max(dataSet[:,j])
        rangeJ = float(maxJ - minJ)
        centroids[:,j] = minJ + rangeJ * np.random.rand(k, 1)
    return centroids
#计算欧几里得距离
def distEclud(vecA, vecB):
    return np.sqrt(np.sum(np.power(vecA - vecB, 2))) # 求两个向量之间的距离
def plot_centroid(data, clusterAssignment):
    m = np.shape(data)[0]
    for i in range(m):
        if clusterAssignment[i,0]==0:
            plt.scatter(data[i, 0], data[i, 1], c = "red", marker='o')
        elif clusterAssignment[i,0]==1:
            plt.scatter(data[i, 0], data[i, 1], c = "green", marker='*')  
        else :
            plt.scatter(data[i, 0], data[i, 1], c = "blue", marker='+') 
    plt.xlabel('petal length')  
    plt.ylabel('petal width')    
    plt.show() 
def kmeans(data, k):#k类
    centroids = randCent(data,k)
    #print (centroids)
    m = np.shape(data)[0]#行数，一共有多少数据
    clusterChanged = True
    clusterAssignment = np.zeros((m,2)) # 存放每个样本点的所属类别和距离
    #当所有点的所属点的类别状态不再变化之后，停止更新
    while clusterChanged:
        clusterChanged = False
        for i in range(m):
            minDist = np.inf#初始化为无限大的数
            minIndex = 0
            #计算当前样本点与centroid的距离
            for j in range(k):#找到离哪个点最近
                dist = distEclud(data[i,:],centroids[j,:])
                if dist < minDist:
                    minDist = dist
                    minIndex = j
            if clusterAssignment[i,0] != minIndex:
                clusterAssignment[i,0] = minIndex
                clusterAssignment[i,1] = minDist
                clusterChanged = True
        print ("-----------------")
        #print (centroids)
        #print (clusterAssignment)   
        # 重新计算中心点
        for cent in range(k):   
            ptsInClust = data[np.nonzero(clusterAssignment[:,0] == cent)[0]]   # 取第一列等于cent的所有行，nonzero返回非零的索引
            #print (np.nonzero(clusterAssignment[:,0] == cent)[0])
            centroids[cent,:] = np.mean(ptsInClust, axis = 0)  # 算出这些数据的中心点
    return centroids, clusterAssignment
if __name__=="__main__":
    data = LoadData('iris.csv')
    c, a = kmeans(data,3)
    print (c)
    plot_centroid(data[:,2:4],a)