獲取m個(gè)n維的數(shù)據(jù)
隨即選取k個(gè)點(diǎn)作為初始中心點(diǎn)
while keep_changing:
	for i in range(m):
		for j in range(k):
			計(jì)算每個(gè)點(diǎn)到center的距離
			判斷離哪個(gè)點(diǎn)更近
		for center in range(k):
			更新類別中心點(diǎn)的坐標(biāo)

import numpy as np
import matplotlib.pyplot as plt
import sklearn.datasets as datasets

def create_data():
    X,y = datasets.make_blobs(n_samples=1000,n_features=2,centers=[[1,0],[5,4],[2,3],[10,8],[7,4]])
    return X,y

def init_centers(data,k):
    m, n =data.shape
    # m 樣本個(gè)數(shù)，n特征個(gè)數(shù)
    center_ids = np.random.choice(m,k)
    centers = data[center_ids]
    return centers

def cal_dist(ptA,ptB):
   return np.linalg.norm(ptA-ptB)

def kmeans_process(data,k):
    centers = init_centers(data, k)
    m, n = data.shape
    keep_changing = True
    pred_y = np.zeros((m,))

    while keep_changing:
        keep_changing = False
        # 計(jì)算剩余樣本所屬類別
        for i in range(m):
            min_distance = np.inf
            for center in range(k):
                distance = cal_dist(data[i,:],centers[center,:])
                if distance<min_distance: # 判斷離哪個(gè)更近
                    min_distance = distance
                    idx = center # 類別換下
            if pred_y[i] != idx:   # 判斷是否發(fā)生了改變
                keep_changing = True
            pred_y[i] = idx
        # 更新類別中心點(diǎn)坐標(biāo)
        for center in range(k):
            cluster_data = data[pred_y==center]
            centers[center,:] = np.mean(cluster_data, axis=0) # 求相同類別數(shù)據(jù)點(diǎn)的質(zhì)心點(diǎn)
        print(centers)
    return centers, pred_y

if __name__ == '__main__':
    X, y = create_data()
    centers , pred_y = kmeans_process(data=X, k=5)
    plt.scatter(X[:,0], X[:,1], s=3, c=pred_y)
    plt.scatter(centers[:,0], centers[:,1], s=10, c='k')
    plt.show()