机器学习--kmeans的基本实现

九点前吃饭

于 2019-06-09 15:42:31 发布

阅读量399

点赞数 1

文章标签：机器学习 kmeans

本文链接：https://blog.youkuaiyun.com/haiqiang1995/article/details/91353598

版权

本文介绍并实现了K-means聚类算法，通过Python和NumPy库完成数据集加载、初始化聚类中心、计算距离、迭代更新聚类中心直至收敛的过程。K-means是一种常用的无监督学习算法，用于将数据集划分为多个簇。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

# _*_ coding:utf-8 _*_
import numpy as np
def loadDataset():
    a = np.array([(3, 4), (3, 6), (7, 3), (4, 7), (3, 8), (8, 5), (4, 5), (4, 1), (7, 4), (5, 5)]).astype(np.float)
    return a

def initCenter(dataset,k):
    index = np.random.choice(len(dataset),k,replace=False)
    # print(dataset[index])
    return dataset[index]

def cal_dis(a,b):
    return np.sum((a-b)**2)**0.5

def kmeans(dataset,k):
    centers = initCenter(dataset,k)
    m = dataset.shape[0]
    clusters = np.full(m,np.nan)
    flag = True
    while(flag):
        flag=False
        for i in range(len(dataset)):
            mini_dst,index = 9999,-1
            for j in range(len(centers)):
                dst = cal_dis(dataset[i],centers[j])
                #判断最小距离是否发生变化
                if dst<mini_dst:
                    mini_dst = dst
                    index = j
            if clusters[i]!=index:
                clusters[i]=index
                flag =True

        #更新centers
        print(np.where(clusters==1))
        for i in range(len(centers)):
            centers[i] = np.mean(dataset[np.where(clusters==i)])
    return clusters,centers



if __name__ == '__main__':
    dataset = loadDataset()
    clusters,centers = kmeans(dataset,3)
    print(centers,clusters)