第120天：机器学习算法之 K 均值聚类

import numpy as npimport pandas as pdimport random
class Clusters():    def __init__(self, train_data, K):        '''        :params train_data: ndarray. 训练数据.        :params K: int. 要划分的簇的数量.        '''        super().__init__()
        self.train_data = train_data        self.K = K        # 标记聚类是否完成。具体的真假，取决于是否还存在需要从一个簇移动到另一个簇的数据        self.finished = False
        # 随机选取 K 个数据作为各个簇的中心点        index = random.sample(range(len(self.train_data)), self.K)        self.centroid = train_data[index, 1:5]
        # 将训练数据均匀分配到各个簇，以便以同一的形式适用于数据的分配        self.clusters = []        offset = len(train_data) // self.K        for i in range(self.K):            start = offset * i            if i < self.K-1:                self.clusters.append(train_data[start:start+offset,:])            else:                # 最后一个簇包含剩下的所有数据                self.clusters.append(train_data[start:,:])                    # 加载所要用到的数据集    @staticmethod    def getData():        '''        获取数据，返回值类型为 ndarray        '''        train_data = pd.read_csv('iris.csv').to_numpy()
        return train_data

# 将各数据分配到每个簇中去    def assign(self):        self.finished = True        # data_index_list 和 target_index_list 分别记录“需要移动的数据在当前簇中的索引”以及“要移动到的目标簇索引”        target_index_list = []        data_index_list = []        for i in range(self.K):            target_index_list.append([])            data_index_list.append([])
        for cluster_index in range(len(self.clusters)):            for data_index in range(len(self.clusters[cluster_index])):                diff = self.clusters[cluster_index][data_index, 1:5] - self.centroid                distance_square = np.sum(diff * diff, axis=1)                target_index = np.argmin(distance_square)
                if cluster_index != target_index:                    self.finished = False                    target_index_list[cluster_index].append(target_index)                    data_index_list[cluster_index].append(data_index)                for cluster_index in range(self.K):            for index in range(len(target_index_list[cluster_index])):                target_index = target_index_list[cluster_index][index]                data_index = data_index_list[cluster_index][index]
                self.clusters[target_index] = np.append(self.clusters[target_index],                                                         self.clusters[cluster_index][data_index, :]).reshape(-1, 6)
        for cluster_index in range(self.K):            data_index = data_index_list[cluster_index]            self.clusters[cluster_index] = np.delete(self.clusters[cluster_index], data_index, axis=0)

# 更新各个簇的质心    def update(self):        for cluster_index in range(len(self.clusters)):            self.centroid[cluster_index] = np.mean(self.clusters[cluster_index][:,1:5], axis=0)

def train(self):        '''        进行聚类训练        '''        while not self.finished:            self.assign()            self.update()        print('训练完成！！！')

def printResult(self):        '''        打印聚类结果        '''        print('-'*80)        print('*'*80)        print('-'*80)        print('*'*30, '聚类结果', '*'*30)        print('-'*30,'各簇中心','-'*30)        for i in range(self.K):            print('第', str(i), '簇中心：', self.centroid[i])        print('-'*80)        print('-'*30,'各簇结果','-'*30)        for i in range(self.K):            print('-'*20, '第', str(i), '簇结果', '-'*20,)            for d in self.clusters[i]:                print(d[5])
        print('-'*80)        print('*'*80)        print('-'*80)

print('-'*80)K = int(input('请输入要划分的簇数（应为正整数）：'))data = Clusters.getData()clusters = Clusters(data, K)clusters.train()clusters.printResult()