python实现kmeans算法

        这里是对鸢尾花进行分类，如果要修改，只需要换路径df还有种类k就可以了。具体逻辑在很多博客讲解已经很清楚了。

import pandas as pd
import numpy as np
import random
import matplotlib.pyplot as plt

df = pd.read_csv("C:/Users/wxc/Desktop/xuexi/pythonProject/机器学习/分类与聚类/kmeans/ris.csv")
X_train = df.iloc[:, 0:4].values# 将数据集转换为numpy数组
k = 3# 设置聚类中心的个数
random_point = X_train[random.sample(range(len(X_train)), k)]# 这里是用的numpy数组，从 X_train 中随机选择 k 个样本作为初始的聚类中心

# 计算欧式距离，并将每个样本分配到最近的聚类中心
def assign_cluster(X_train, random_point):
    cluster_assignment = []
    for x in X_train:
        distances = [np.linalg.norm(x - point) for point in random_point] #这里在进行欧式距离的计算
        closest_cluster = np.argmin(distances)  # 获取最小距离对应的聚类中心的索引值
        cluster_assignment.append(closest_cluster) #添加进去
    return cluster_assignment

# 更新聚类中心
def New_centers(X_train, cluster_assignment, k):
    new_centers = []
    for i in range(k):
        cluster_mean = X_train[cluster_assignment == i].mean(axis=0) # 求取均值，这样子就能更新聚类中心了
        new_centers.append(cluster_mean)
    return new_centers

# kmeans算法
def kmeans(X_train, random_point, k):
    for i in range(100):  # 这个迭代次数可以修改
        cluster_assignment = assign_cluster(X_train, random_point)
        new_centers = New_centers(X_train, np.array(cluster_assignment), k)
        if np.allclose(random_point, new_centers, rtol=1e-05):  # 由于是nd.array类型，无法直接使用==，所以这里做差表示
            break

        random_point = new_centers
    return cluster_assignment, new_centers

cluster_assignment, final_centers = kmeans(X_train, random_point, k)
final_centers = np.array(final_centers)

for i in range(len(X_train)):
    print(f"样本{i+1}的具体分类: {cluster_assignment[i]}")

print("最终的中心点位置:", final_centers)# 输出每个样本所属的具体分类


# 绘制分类的图像
fig, ax = plt.subplots()
colors = ['red', 'green', 'blue']  # 设置每个聚类的颜色

for i in range(k):
    cluster_samples = [X_train[j] for j in range(len(X_train)) if cluster_assignment[j] == i] # 遍历一遍所属类别的元素
    cluster_samples = np.array(cluster_samples) #先变成np.array类型
    ax.scatter(cluster_samples[:, 0], cluster_samples[:, 1], c=colors[i], label=f'Cluster {i+1}')  #以第一特征为x轴，第二特征为y轴，开始绘制

ax.scatter(final_centers[:, 0], final_centers[:, 1], c='black', marker='x', label='Centroids')
ax.legend()
plt.show()