如何使用 python 直观地比较集群?

How to visually compare clusters using python?

我正在研究用于客户细分的 k 均值聚类。我的输入数据有 12 个特征和 7315 行。

因此,我尝试了下面的代码来执行 k-means

kmeans = KMeans(n_clusters = 5, init = "k-means++", random_state = 42)
data_normalized['y_kmeans'] = kmeans.fit_predict(data_normalized)

为了可视化,我尝试了下面的代码

u_labels = np.unique(data_normalized['y_kmeans'])
 
#plotting the results:
 
for i in u_labels:
    plt.scatter(data_normalized[y_kmeans == i , 0] , data_normalized[y_kmeans == i , 1] , label = i)
plt.legend()
plt.show()

我收到如下错误

TypeError: '(array([False, False, False, ..., False, False, False]), 0)' is an invalid key

InvalidIndexError: (array([False, False, False, ..., False, False, False]), 0)

如何可视化我的集群以查看它们彼此之间的距离?

由于我没有你的数据集,我模拟了你的数据框如下: (我假设有 9 个不同的集群组)

d={'col1': [i/100 for i in random.choices(range(1,100), k=7315)],
       'col2':[i/100 for i in random.choices(range(1,100), k=7315)],
       'y_kmeans':random.choices(range(1,10), k=7315)}
data_normalized = pd.DataFrame(d)

之后您可以按如下方式绘制集群,

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

u_labels = np.unique(data_normalized['y_kmeans']).tolist()

scatter = plt.scatter(data_normalized['col1'], data_normalized['col2'],
            c=data_normalized['y_kmeans'], cmap='tab20')
plt.legend(handles=scatter.legend_elements()[0], labels=u_labels)
plt.show()

我得到以下聚类图