如何使用Matplotlib从多功能kmeans模型绘制聚类和中心？ 功能0 功能1 功能2 功能3 功能4 功能5 功能6

如何解决如何使用Matplotlib从多功能kmeans模型绘制聚类和中心？ 功能0 功能1 功能2 功能3 功能4 功能5 功能6

我使用kmeans算法来确定数据集中的簇数。在下面的代码中，您可以看到我具有多种功能，有些是绝对的，有些则没有。我对它们进行了编码和缩放，得到了最佳的簇数。

您可以从此处下载数据： https://www.sendspace.com/file/1cnbji

import sklearn.metrics as sm

from sklearn.preprocessing import scale

from sklearn.preprocessing import Normalizer
from sklearn.preprocessing import StandardScaler,MinMaxScaler

from sklearn.cluster import KMeans,SpectralClustering,MiniBatchKMeans
from sklearn.compose import ColumnTransformer
from sklearn.preprocessing import OneHotEncoder

import matplotlib.pyplot as plt

import pandas as pd



df = pd.read_csv('dataset.csv')
print(df.columns)

features = df[['parcela','bruto','neto','osnova','sipovi','nadzemno','podzemno','tavanica','fasada']]

trans = ColumnTransformer(transformers=[('onehot',OneHotEncoder(),['tavanica','fasada']),('StandardScaler',Normalizer(),['parcela','sipovi'])],remainder='passthrough') # Default is to drop untransformed columns

features = trans.fit_transform(features)

Sum_of_squared_distances = []
for i in range(1,19):

     kmeans = KMeans(n_clusters = i,init = 'k-means++',random_state = 0)
     kmeans.fit(features)
     Sum_of_squared_distances.append(kmeans.inertia_)


plt.plot(range(1,19),Sum_of_squared_distances,'bx-')
plt.xlabel('k')
plt.ylabel('Sum_of_squared_distances')
plt.title('Elbow Method For Optimal k')
plt.show()

• 在图中，肘部方法显示我的最佳簇数为7。
• 如何绘制7个群集？
  • 我想在图形上看到质心，并用7种不同颜色的群集散布图。

解决方法

• 给出Plot: kmeans clustering centroid，其中centers是一维。 centers数组的形状为(3,2)，其中x为(3,1)，y为(3,1)。
  • 针对此一维中心展示的方法已被修改为该问题的模型所产生的针对七个维中心的解决方案。
• 此问题中为模型返回的centers具有七个维度，形状为(7,14)，其中14是7组x和y值。
• 此解决方案回答了以下问题：如何绘制聚类和中心？
  • 它不提供对模型结果的评论或解释，这可能是在SE: Cross Validated或SE: Data Science中提出的另一个问题。

# uses the imports as shown in the question
from matplotlib.patches import Rectangle,Patch  # for creating a legend
from matplotlib.lines import Line2D

# beginning with 
features = trans.fit_transform(features)

# create the model and fit it to features
kmeans_model2 = KMeans(n_clusters=7,init='k-means++',random_state=0).fit(features)

# find the centers; there are 7
centers = np.array(kmeans_model2.cluster_centers_)

# unique markers for the labels
markers = ['o','v','s','*','p','d','h']

# get the model labels
labels = kmeans_model2.labels_
labels_unique = set(labels)

# unique colors for each label
colors = sns.color_palette('husl',n_colors=len(labels_unique))

# color map with labels and colors
cmap = dict(zip(labels_unique,colors))

# plot
# iterate through each group of 2 centers
for j in range(0,len(centers)*2,2):
    plt.figure(figsize=(6,6))
    
    x_features = features[:,j]
    y_features = features[:,j+1]
    x_centers = centers[:,j]
    y_centers = centers[:,j+1]
    
    # add the data for each label to the plot
    for i,l in enumerate(labels):
#         print(f'Label: {l}')  # uncomment as needed
#         print(f'feature x coordinates for label:\n{x_features[i]}')  # uncomment as needed
#         print(f'feature y coordinates for label:\n{y_features[i]}')  # uncomment as needed
        plt.plot(x_features[i],y_features[i],color=colors[l],marker=markers[l],alpha=0.5)

    # print values for given plot,rounded for easier interpretation; all 4 can be commented out
    print(f'feature labels:\n{list(labels)}')
    print(f'x_features:\n{list(map(lambda x: round(x,3),x_features))}')
    print(f'y_features:\n{list(map(lambda x: round(x,y_features))}')
    print(f'x_centers:\n{list(map(lambda x: round(x,x_centers))}')
    print(f'y_centers:\n{list(map(lambda x: round(x,y_centers))}')
    
    # add the centers
    # this loop is to color the center marker to correspond to the color of the corresponding label.
    for k in range(len(centers)):  
        plt.scatter(x_centers[k],y_centers[k],marker="X",color=colors[k])
    
    # title
    plt.title(f'Features: Dimension {int(j/2)}')
    
    # create the rectangles for the legend
    patches = [Patch(color=v,label=k) for k,v in cmap.items()]
    # create centers marker for the legend
    black_x = Line2D([],[],color='k',marker='X',linestyle='None',label='centers',markersize=10)
    # add the legend
    plt.legend(title='Labels',handles=patches + [black_x],bbox_to_anchor=(1.04,0.5),loc='center left',borderaxespad=0,fontsize=15)
    
    plt.show()

绘图输出

• 许多绘制的要素具有重叠的值和中心。
• 已打印x和y的{​​{1}}和features值，以更容易地看到重叠并确认绘制的值。
  • 负责的centers行可以在不再需要时被注释掉或删除。

功能0

print

功能1

feature labels:
[6,1,5,3,4,2,6,6]
x_features:
[0.0,1.0,0.0,0.0]
y_features:
[1.0,1.0]
x_centers:
[1.0,0.0]
y_centers:
[0.0,-0.0,1.0]

功能2

feature labels:
[6,1.0]

功能3

feature labels:
[6,0.0]
y_features:
[0.0,0.0]
x_centers:
[0.0,0.125,0.0]

功能4

feature labels:
[6,0.0]
y_features:
[0.298,0.193,0.18,0.336,0.181,0.174,0.197,0.23,0.175,0.212,0.196,0.186,0.2,0.15,0.141,0.304,0.108,0.101,0.105,0.459,0.16,0.224,0.216,0.246,0.139,0.111,0.227,0.177,0.159,0.25,0.298,0.223,0.335,0.431,0.17,0.381,0.255,0.222,0.296,0.156,0.202,0.145,0.195,0.336]
x_centers:
[0.0,0.875,0.0]
y_centers:
[0.196,0.188,0.249,0.237,0.182,0.328]

功能5

feature labels:
[6,6]
x_features:
[0.712,0.741,0.763,0.704,0.749,0.754,0.735,0.744,0.738,0.743,0.747,0.758,0.759,0.714,0.766,0.748,0.728,0.755,0.681,0.752,0.762,0.734,0.721,0.756,0.737,0.742,0.724,0.712,0.733,0.73,0.688,0.722,0.705,0.777,0.764,0.739,0.76,0.704]
y_features:
[0.614,0.636,0.612,0.601,0.631,0.64,0.62,0.624,0.633,0.632,0.63,0.61,0.629,0.641,0.616,0.65,0.644,0.539,0.628,0.623,0.627,0.603,0.648,0.614,0.58,0.562,0.666,0.587,0.565,0.591,0.646,0.642,0.625,0.601]
x_centers:
[0.745,0.708]
y_centers:
[0.63,0.611,0.604]

功能6

feature labels:
[6,6]
x_features:
[0.164,0.096,0.103,0.171,0.091,0.106,0.094,0.132,0.098,0.102,0.115,0.079,0.095,0.135,0.075,0.088,0.126,0.063,0.134,0.107,0.09,0.072,0.097,0.073,0.123,0.165,0.154,0.133,0.158,0.084,0.11,0.1,0.164,0.069,0.171]
y_features:
[0.001,0.002,0.001,0.005,0.004,0.003,0.001]
x_centers:
[0.093,0.116,0.112,0.152]
y_centers:
[0.001,0.001]

在一个图上更新所有尺寸

• 根据OP的要求

feature labels:
[6,0.0]

• 正如各个地块所指出的，有很多重叠之处。