1+ from sklearn import datasets
2+ import matplotlib .pyplot as plt
3+ import numpy as np
4+
5+ iris = datasets .load_iris ()
6+ X , y = iris .data , iris .target
7+
8+ # 为了便于可视化，只取两个维度
9+ data = X [:,[1 ,3 ]]
10+
11+ print (data )
12+
13+ plt .scatter (data [:,0 ],data [:,1 ])
14+
15+ ck = 3
16+ '''
17+ 随机选取k个点为聚类的初始代表点，即质点
18+ '''
19+ def rand_center (data ,k ):
20+ """Generate k center within the range of data set."""
21+ n = data .shape [1 ] # features
22+ centroids = np .zeros ((k ,n )) # init with (0,0)....
23+ for i in range (n ):
24+ dmin , dmax = np .min (data [:,i ]), np .max (data [:,i ])
25+ centroids [:,i ] = dmin + (dmax - dmin ) * np .random .rand (k )
26+ return centroids
27+
28+ # 初始化点列表
29+ centroids = rand_center (data , ck )
30+ print (centroids )
31+
32+ def kmeans (data ,k = 2 ):
33+ def _distance (p1 ,p2 ):
34+ """
35+ Return Eclud distance between two points.
36+ p1 = np.array([0,0]), p2 = np.array([1,1]) => 1.414
37+ """
38+ tmp = np .sum ((p1 - p2 )** 2 )
39+ return np .sqrt (tmp )
40+ def _rand_center (data ,k ):
41+ """Generate k center within the range of data set."""
42+ n = data .shape [1 ] # features
43+ centroids = np .zeros ((k ,n )) # init with (0,0)....
44+ for i in range (n ):
45+ dmin , dmax = np .min (data [:,i ]), np .max (data [:,i ])
46+ centroids [:,i ] = dmin + (dmax - dmin ) * np .random .rand (k )
47+ return centroids
48+
49+ def _converged (centroids1 , centroids2 ):
50+
51+ # if centroids not changed, we say 'converged'
52+ set1 = set ([tuple (c ) for c in centroids1 ])
53+ set2 = set ([tuple (c ) for c in centroids2 ])
54+ return (set1 == set2 )
55+
56+
57+ n = data .shape [0 ] # number of entries
58+ centroids = _rand_center (data ,k )
59+ label = np .zeros (n ,dtype = np .int ) # track the nearest centroid
60+ assement = np .zeros (n ) # for the assement of our model
61+ converged = False
62+
63+ while not converged :
64+ old_centroids = np .copy (centroids )
65+ for i in range (n ):
66+ # determine the nearest centroid and track it with label
67+ min_dist , min_index = np .inf , - 1
68+ for j in range (k ):
69+ dist = _distance (data [i ],centroids [j ])
70+ if dist < min_dist :
71+ min_dist , min_index = dist , j
72+ label [i ] = j
73+ assement [i ] = _distance (data [i ],centroids [label [i ]])** 2
74+
75+ # update centroid
76+ for m in range (k ):
77+ centroids [m ] = np .mean (data [label == m ],axis = 0 )
78+ converged = _converged (old_centroids ,centroids )
79+ return centroids , label , np .sum (assement )
80+
81+
82+ # 多运行
83+ best_assement = np .inf
84+ best_centroids = None
85+ best_label = None
86+
87+ for i in range (10 ):
88+ centroids , label , assement = kmeans (data ,ck )
89+ if assement < best_assement :
90+ best_assement = assement
91+ best_centroids = centroids
92+ best_label = label
93+
94+ data0 = data [best_label == 0 ]
95+ data1 = data [best_label == 1 ]
96+
97+ # 打印展示
98+ fig , (ax1 ,ax2 ) = plt .subplots (1 ,2 ,figsize = (12 ,5 ))
99+ ax1 .scatter (data [:,0 ],data [:,1 ],c = 'c' ,s = 30 ,marker = 'o' )
100+ ax2 .scatter (data0 [:,0 ],data0 [:,1 ],c = 'r' )
101+ ax2 .scatter (data1 [:,0 ],data1 [:,1 ],c = 'c' )
102+ ax2 .scatter (centroids [:,0 ],centroids [:,1 ],c = 'b' ,s = 120 ,marker = 'o' )
103+ plt .show ()
0 commit comments