学习sklearn聚类使用

  1 #!/usr/bin/python
  2 # -*- coding: utf-8 -*-
  3 
  4 import numpy as np
  5 import matplotlib.pyplot as plt
  6 from sklearn.cluster import KMeans
  7 from sklearn.cluster import MeanShift
  8 from sklearn.metrics import homogeneity_completeness_v_measure
  9 from sklearn import mixture
 10 from scipy.cluster.hierarchy import linkage
 11 from scipy.cluster.hierarchy import fcluster
 12 
 13 class ClusterMethod:
 14 
 15     def __init__(self):
 16         l1=np.zeros(100)
 17         l2=np.ones(100)
 18         self.labels=np.concatenate((l1,l2),)
 19 
 20     #随机创建两个二维正太分布，形成数据集
 21     def dataProduction(self):
 22         # 随机创建两个二维正太分布，形成数据集
 23         np.random.seed(4711)
 24         c1 = np.random.multivariate_normal([10, 0], [[3, 1], [1, 4]], size=[100, ])
 25         l1 = np.zeros(100)
 26         l2 = np.ones(100)
 27         # 一个100行的服从正态分布的二维数组
 28         c2 = np.random.multivariate_normal([0, 10], [[3, 1], [1, 4]], size=[100, ])
 29         # 加上一些噪音
 30         np.random.seed(1)
 31         noise1x = np.random.normal(0, 2, 100)
 32         noise1y = np.random.normal(0, 8, 100)
 33         noise2 = np.random.normal(0, 8, 100)
 34         c1[:, 0] += noise1x  # 第0列加入噪音数据
 35         c1[:, 1] += noise1y
 36         c2[:, 1] += noise2
 37 
 38         # 定义绘图
 39         self.fig = plt.figure(figsize=(20, 15))
 40         # 添加子图，返回Axes实例，参数：子图总行数，子图总列数，子图位置
 41         ax = self.fig.add_subplot(111)
 42         # x轴
 43         ax.set_xlabel(‘x‘, fontsize=30)
 44         # y轴
 45         ax.set_ylabel(‘y‘, fontsize=30)
 46         # 标题
 47         self.fig.suptitle(‘classes‘, fontsize=30)
 48         # 连接
 49         labels = np.concatenate((l1, l2), )
 50         X = np.concatenate((c1, c2), )
 51         # 散点图
 52         pp1 = ax.scatter(c1[:, 0], c1[:, 1], cmap=‘prism‘, s=50, color=‘r‘)
 53         pp2 = ax.scatter(c2[:, 0], c2[:, 1], cmap=‘prism‘, s=50, color=‘g‘)
 54         ax.legend((pp1, pp2), (‘class 1‘, ‘class 2‘), fontsize=35)
 55         self.fig.savefig(‘scatter.png‘)
 56         return X
 57 
 58     def clusterMethods(self):
 59         X=self.dataProduction()
 60         self.fig.clf()#reset plt
 61         self.fig,((axis1,axis2),(axis3,axis4))=plt.subplots(2,2,sharex=‘col‘,sharey=‘row‘)#函数返回一个figure图像和一个子图ax的array列表
 62 
 63         #k-means
 64         self.kMeans(X,axis1)
 65         #mean-shift
 66         self.meanShift(X,axis2)
 67         #gaussianMix
 68         self.gaussianMix(X,axis3)
 69         #hierarchicalWard
 70         self.hierarchicalWard(X,axis4)
 71 
 72     def kMeans(self,X,axis1):
 73         kmeans=KMeans(n_clusters=2)#聚类个数
 74         kmeans.fit(X)#训练
 75         pred_kmeans=kmeans.labels_#每个样本所属的类
 76         print(‘kmeans:‘,np.unique(kmeans.labels_))
 77         print(‘kmeans:‘,homogeneity_completeness_v_measure(self.labels,pred_kmeans))#评估方法，同质性，完整性，两者的调和平均
 78         #plt.scatter(X[:,0],X[:,1],c=kmeans.labels_,cmap=‘prism‘)
 79         axis1.scatter(X[:,0],X[:,1],c=kmeans.labels_,cmap=‘prism‘)
 80         axis1.set_ylabel(‘y‘,fontsize=40)
 81         axis1.set_title(‘k-means‘,fontsize=40)
 82         #plt.show()
 83 
 84     def meanShift(self,X,axis2):
 85         ms=MeanShift(bandwidth=7)#带宽
 86         ms.fit(X)
 87         pred_ms=ms.labels_
 88         axis2.scatter(X[:,0],X[:,1],c=pred_ms,cmap=‘prism‘)
 89         axis2.set_title(‘mean-shift‘,fontsize=40)
 90         print(‘mean-shift:‘,np.unique(ms.labels_))
 91         print(‘mean-shift:‘,homogeneity_completeness_v_measure(self.labels,pred_ms))
 92 
 93     def gaussianMix(self,X,axis3):
 94         gmm=mixture.GMM(n_components=2)
 95         gmm.fit(X)
 96         pred_gmm=gmm.predict(X)
 97         axis3.scatter(X[:, 0], X[:, 1], c=pred_gmm, cmap=‘prism‘)
 98         axis3.set_xlabel(‘x‘, fontsize=40)
 99         axis3.set_ylabel(‘y‘, fontsize=40)
100         axis3.set_title(‘gaussian mixture‘, fontsize=40)
101         print(‘gmm:‘,np.unique(pred_gmm))
102         print(‘gmm:‘,homogeneity_completeness_v_measure(self.labels,pred_gmm))
103 
104     def hierarchicalWard(self,X,axis4):
105         ward=linkage(X,‘ward‘)#训练
106         max_d=110#终止层次算法最大的连接距离
107         pred_h=fcluster(ward,max_d,criterion=‘distance‘)#预测属于哪个类
108         axis4.scatter(X[:,0], X[:,1], c=pred_h, cmap=‘prism‘)
109         axis4.set_xlabel(‘x‘,fontsize=40)
110         axis4.set_title(‘hierarchical ward‘,fontsize=40)
111         print(‘ward:‘,np.unique(pred_h))
112         print(‘ward:‘,homogeneity_completeness_v_measure(self.labels,pred_h))
113 
114         self.fig.set_size_inches(18.5,10.5)
115         self.fig.savefig(‘comp_clustering.png‘,dpi=100)#保存图
116 
117 if __name__==‘__main__‘:
118     cluster=ClusterMethod()
119     cluster.clusterMethods()