-
Notifications
You must be signed in to change notification settings - Fork 2
/
聚类混淆度算法
441 lines (385 loc) · 21.4 KB
/
聚类混淆度算法
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
#author:MingYu Pang
#time：2017/6/23
#下面是一个类，通过输入数据集，聚类数，能得到5种算法的聚类数混淆情况。
#另外，在默认参数条件下，
#能实现DBSCAN以及Affinity Propagation算法的聚类混淆情况（这种情况，无须输入聚类数）
class DrawCluster():
def __init__(self,dataset,cluster_num=3):
self.dataset=dataset #数据集，dataset形式为pd
self.cluster_num=cluster_num #聚类数
self.suanfa={}
self.suanfa_PCA={} #降维后的字典，储存算法聚类结果
self.key=[] #存储算法的名称,不包括DBSCAN以及AffinityPropogation
self.keys=[] #储存算法的名称，包括DBSCAN以及AffinityPropogation
'''可通过下面的变量修改聚类算法的参数 '''
self.KMeans={'n_clusters':self.cluster_num,'init':'k-means++',
'compute_labels':True,'max_iter':100,'batch_size':100,
'tol':0.0,'max_no_improvement':10,'random_state':None,'verbose':0,
'init_size':None, 'n_init':3,
'reassignment_ratio':0.01}
self.AffinityPropagation={'damping':0.5,'max_iter':200, 'convergence_iter':15,
'copy':True, 'preference':None, 'affinity':'euclidean', 'verbose':False}
self.SpectralClustering={'n_clusters':self.cluster_num, 'eigen_solver':None, 'random_state':None,
'n_init':10, 'gamma':1.0, 'affinity':'rbf', 'n_neighbors':10,
'eigen_tol':0.0, 'assign_labels':'kmeans', 'degree':3,
'coef0':1, 'kernel_params':None, 'n_jobs':1}
self.ward={'n_clusters':self.cluster_num, 'affinity':'euclidean',
'connectivity':None, 'compute_full_tree':'auto',
'linkage':'ward', 'pooling_func':np.mean}
self.AgglomerativeClustering={'n_clusters':self.cluster_num, 'affinity':'euclidean',
'connectivity':None, 'compute_full_tree':'auto',
'linkage':'average', 'pooling_func':np.mean}
self.DBSCAN={'eps':0.5, 'min_samples':5, 'metric':'euclidean',
'algorithm':'auto', 'leaf_size':30, 'p':None, 'n_jobs':1}
self.Birch={'threshold':0.5, 'branching_factor':50,
'n_clusters':self.cluster_num, 'compute_labels':True, 'copy':True}
def average(self,df,i,g): #计算聚类中心到聚类样本点的平均距离，i代表第i个算法，g代表聚类数
h={}
if self.key[i]!='DBSCAN' and self.key[i]!='AffinityPropagation':
ds=[] #平均距离
for j in range(g):
#开始计算平均距离
lei=np.nonzero(self.suanfa[self.key[i]]['y_pred']==j)[0]
if len(lei)==0:print(g)
d=0
for k in lei:
d=d+self.enlid(self.suanfa[self.key[i]]['centroids'][j],df.iloc[k,:])
ds.append(d/len(lei))
h=ds
return h
#类似于求平均距离，我们求最大距离以及最小距离作为半径,i为第i个算法，t为聚类数
def maxRminR1(self,df,i,t):
h1=[]
h0=[]
if self.key[i]!='DBSCAN' and self.key[i]!='AffinityPropagation':
dm=[] #最大距离
di=[]#最小距离
for j in range(t):
#开始计算平均距离
lei=np.nonzero(self.suanfa[self.key[i]]['y_pred']==j)[0]
maxd=0
minr=np.inf
for k in lei:
d=self.enlid(self.suanfa[self.key[i]]['centroids'][j],df.iloc[k,:])
if maxd<d:maxd=d
if minr>d:minr=d
dm.append(maxd)
di.append(minr)
h1=dm
h0=di
return h1,h0
def enlid(self,inA,inB): #求解欧氏距离
dist = np.linalg.norm(inA - inB)
return dist
#用于不对应于同算法画圆的函数，k是聚类数，H,Max,Min分别为圆的平均半径，最大半径，最小半径
#平均半径，最大半径，最小半径分别由blue,black,yellow颜色绘制
def plot_Circle(self,k,H,Max,Min):
fig = plt.figure(figsize=(12,8))
for i in range(len(self.key)):
ax = fig.add_subplot(331+i)
'''实现平均距离'''
cirH=0
cirMax=0
cirMin=0
print("%s algorithm :Cluster number,Cluster shape after PCA:"%self.key[i])
print(k,self.suanfa1[self.key[i]]['centroids'].shape)
for t in range(k):
#print(H[key[i]])
#print(suanfa1[key[i]]['centroids'])
#第一个参数为圆心坐标，第二个为半径 #第三个为透明度（0-1）
cirH=Circle(xy = (self.suanfa1[self.key[i]]['centroids'][t][0],self.suanfa1[self.key[i]]['centroids'][t][1]),
radius=H[self.key[i]][0][t], alpha=0.4)
ax.add_patch(cirH)
cirMax=Circle(xy = (self.suanfa1[self.key[i]]['centroids'][t][0],self.suanfa1[self.key[i]]['centroids'][t][1]),
radius=Max[self.key[i]][0][t],color="black",alpha=0.4)
ax.add_patch(cirMax)
cirMin=Circle(xy = (self.suanfa1[self.key[i]]['centroids'][t][0],self.suanfa1[self.key[i]]['centroids'][t][1]),
radius=Min[self.key[i]][0][t],color="yellow", alpha=0.4)
ax.add_patch(cirMin)
plt.scatter(self.suanfa1[self.key[i]]['centroids'][t,0],
self.suanfa1[self.key[i]]['centroids'][t,1],marker='8',color="red")
plt.axis('scaled')
plt.axis('equal')
plt.title(self.key[i])
plt.show()
def plot_Circle_special(self,H,Max,Min):
fig = plt.figure(figsize=(12,8))
key=['DBSCAN','AffinityPropagation']
#print('blue--Average distance;black--Max distance;yellow--Min distance;')
for i in range(2):
print("%s algorithm :Cluster number,Cluster shape after PCA:"%key[i])
print(len(self.suanfa[key[i]]['centroids']),self.suanfa1[key[i]]['centroids'].shape)
ax = fig.add_subplot(332+i)
'''实现平均距离'''
cirH=0
cirMax=0
cirMin=0
p=len(self.suanfa[key[i]]['centroids'])
for t in range(p):
#print(H[key[i]])
#print(suanfa1[key[i]]['centroids'])
#第一个参数为圆心坐标，第二个为半径 #第三个为透明度（0-1）
cirH=Circle(xy = (self.suanfa1[key[i]]['centroids'][t][0],self.suanfa1[key[i]]['centroids'][t][1]),
radius=H[key[i]][0][t], alpha=0.4)
ax.add_patch(cirH)
cirMax=Circle(xy = (self.suanfa1[key[i]]['centroids'][t][0],self.suanfa1[key[i]]['centroids'][t][1]),
radius=Max[key[i]][0][t],color="black",alpha=0.4)
ax.add_patch(cirMax)
cirMin=Circle(xy = (self.suanfa1[key[i]]['centroids'][t][0],self.suanfa1[key[i]]['centroids'][t][1]),
radius=Min[key[i]][0][t],color="yellow", alpha=0.4)
ax.add_patch(cirMin)
plt.scatter(self.suanfa1[key[i]]['centroids'][t,0],
self.suanfa1[key[i]]['centroids'][t,1],marker='8',color="red")
plt.axis('scaled')
plt.axis('equal')
plt.title(key[i])
plt.show()
#衡量各种算法的聚类个数均衡度，返回值为每种算法各个聚类的样本个数 (包括DBSCAN，AffinityPropagation)
def Jun_level(self):
keys=list(self.suanfa.keys())
D={}
for j in range(len(self.suanfa)): #多少种算法
if keys[j]!='DBSCAN' and keys[j]!='AffinityPropagation':
line=[]
m=len(set(self.suanfa['MiniBatchKMeans']['y_pred'])) #多少聚类数
for i in range(m):
p=np.nonzero(self.suanfa[keys[j]]['y_pred']==i)[0]
line.append(len(p))
D[keys[j]]=line
else:
lei=list(set(self.suanfa[keys[j]]['y_pred']))
line=[]
for i in lei:
p=np.nonzero(self.suanfa[keys[j]]['y_pred']==i)[0]
line.append(len(p))
D[keys[j]]=line
print('聚类均衡性\n')
print(D)
def special_algorithm(self): #单独考虑DBSCAN以及Affinity算法
dbscan = cluster.DBSCAN(eps=self.DBSCAN['eps'], min_samples=self.DBSCAN['min_samples'],
metric=self.DBSCAN['metric'], algorithm=self.DBSCAN['algorithm'],
leaf_size=self.DBSCAN['leaf_size'], p=self.DBSCAN['p'], n_jobs=self.DBSCAN['n_jobs'])
affinity_propagation = cluster.AffinityPropagation(damping=self.AffinityPropagation['damping'],
max_iter=self.AffinityPropagation['max_iter'],
convergence_iter=self.AffinityPropagation['convergence_iter'],
copy=self.AffinityPropagation['copy'],
preference=self.AffinityPropagation['preference'],
affinity=self.AffinityPropagation['affinity'],
verbose=self.AffinityPropagation['verbose'])
clustering_algorithms = [dbscan,affinity_propagation]
clustering_names = ['DBSCAN','AffinityPropagation']
for name, algorithm in zip(clustering_names, clustering_algorithms):
self.suanfa[name]={}
t0 = time.time()
algorithm.fit(dataset)
t1 = time.time()
self.suanfa[name]['time']=t1-t0
if hasattr(algorithm, 'labels_'):
y_pred = algorithm.labels_.astype(np.int)
else:
y_pred = algorithm.predict(dataset)
l=len(set(y_pred))
y_pred=np.array([y_pred[l] if y_pred[l]!=-1 else l-1 for l in range(len(y_pred))])
self.suanfa[name]['y_pred']=y_pred
if hasattr(algorithm, 'cluster_centers_'):
centers = algorithm.cluster_centers_
self.suanfa[name]['centroids']=centers
else:
#如果算法没有centers_属性，则采用聚类数求平均值方法。
l=len(set(self.suanfa[name]['y_pred']))
le=set(self.suanfa[name]['y_pred'])
p=[];c=np.zeros((l,self.dataset.shape[1]))
t=0
for j in le:
p=np.nonzero(self.suanfa[name]['y_pred']==j)[0]
data=np.mat(self.dataset)[p,:]
if data.shape[0]!=1:
c[t]=np.mean(data,axis=0)
else:
c[t]=data
t=t+1
self.suanfa[name]['centroids']=np.array(c)
key=list(self.suanfa.keys())
self.keys=key
suanfa1= copy.deepcopy(self.suanfa) #suanfa1储存pca后的聚类中心
H={} #用于储存每个聚类的半径值(平均半径)
Max={} #最大距离
Min={}#最小距离
key=['DBSCAN','AffinityPropagation']
num1=len(self.suanfa['DBSCAN']['centroids'])
num2=len(self.suanfa['AffinityPropagation']['centroids'])
for j in range(2):
H[key[j]]=[]
Max[key[j]]=[]
Min[key[j]]=[]
if key[j]=='DBSCAN' :
df=pd.concat([dataset,pd.DataFrame(self.suanfa['DBSCAN']['centroids'])])
redu=PCA(n_components=2).fit_transform(df)
suanfa1[key[j]]['centroids']=redu[-num1:,:]
h=self.average_special(df[:-num1],j,self.suanfa['DBSCAN']['y_pred']) #计算平均距离，j是算法索引，num1是聚类数
H[key[j]].append(h)
M1,M0=self.maxRminR1_special(df[:-num1],j,self.suanfa['DBSCAN']['y_pred'])
Max[key[j]].append(M1)
Min[key[j]].append(M0)
else:
df=pd.concat([dataset,pd.DataFrame(self.suanfa['AffinityPropagation']['centroids'])])
redu=PCA(n_components=2).fit_transform(df)
suanfa1[key[j]]['centroids']=redu[-num2:,:]
h=self.average_special(df[:-num2],j,self.suanfa['AffinityPropagation']['y_pred']) #计算平均距离，j是算法索引，num2是聚类数
H[key[j]].append(h)
#print(H[key[j]])
M1,M0=self.maxRminR1_special(df[:-num2],j,self.suanfa['AffinityPropagation']['y_pred'])
Max[key[j]].append(M1)
Min[key[j]].append(M0)
#print(Min[key[j]])
self.suanfa1=suanfa1
self.plot_Circle_special(H,Max,Min) #绘图函数，聚类数i
self.Jun_level()
def average_special(self,df,i,y_pred):
key=['DBSCAN','AffinityPropagation']
if key[i]=='DBSCAN' or key[i]=='AffinityPropagation':
ds=[] #平均距离
count=0 #用于遍历聚类中心
p=list(set(y_pred))
for j in p:
#开始计算平均距离
lei=np.nonzero(self.suanfa[key[i]]['y_pred']==j)[0]
d=0
for k in lei:
d=d+self.enlid(self.suanfa[key[i]]['centroids'][count],df.iloc[k,:])
count=count+1
ds.append(d/len(lei))
h=ds
return h
def maxRminR1_special(self,df,i,y_pred):
h1=[]
h0=[]
key=['DBSCAN','AffinityPropagation']
if key[i]=='DBSCAN' or key[i]=='AffinityPropagation':
dm=[] #最大距离
di=[]#最小距离
p=list(set(y_pred)) #聚类的标号
count=0
for j in p:
#开始计算平均距离
lei=np.nonzero(self.suanfa[self.key[i]]['y_pred']==j)[0]
maxd=0
minr=np.inf
for k in lei:
d=self.enlid(self.suanfa[key[i]]['centroids'][count],df.iloc[k,:])
if maxd<d:maxd=d
if minr>d:minr=d
count+=1
dm.append(maxd)
di.append(minr)
h1=dm
h0=di
return h1,h0
def Drawnet(self):
'''connectivity 的计算方法，可以用于AgglomerativeClustering以及ward'''
#connectivity = kneighbors_graph(dataset, n_neighbors=7, include_self=False)
#connectivity = 0.5 * (connectivity + connectivity.T)
two_means = cluster.MiniBatchKMeans(n_clusters=self.cluster_num, init=self.KMeans['init'], max_iter=self.KMeans['max_iter'],
batch_size=self.KMeans['batch_size'],verbose=self.KMeans['verbose'],
compute_labels=self.KMeans['compute_labels'],
random_state=self.KMeans['random_state'], tol=self.KMeans['tol'],
max_no_improvement=self.KMeans['max_no_improvement'], init_size=self.KMeans['init_size'],
n_init=self.KMeans['n_init'], reassignment_ratio=self.KMeans['reassignment_ratio'])
ward = cluster.AgglomerativeClustering(n_clusters=self.cluster_num, affinity=self.ward['affinity'],
connectivity=self.ward['connectivity'],
compute_full_tree=self.ward['compute_full_tree'],
linkage=self.ward['linkage'], pooling_func=self.ward['pooling_func'])
spectral = cluster.SpectralClustering(n_clusters=self.cluster_num, eigen_solver=self.SpectralClustering['eigen_solver'],
random_state=self.SpectralClustering['random_state'],
n_init=self.SpectralClustering['n_init'], gamma=self.SpectralClustering['gamma'],
affinity=self.SpectralClustering['affinity'], n_neighbors=self.SpectralClustering['n_neighbors'],
eigen_tol=self.SpectralClustering['eigen_tol'],
assign_labels=self.SpectralClustering['assign_labels'], degree=self.SpectralClustering['degree'],
coef0=self.SpectralClustering['coef0'], kernel_params=self.SpectralClustering['kernel_params'],
n_jobs=self.SpectralClustering['n_jobs'])
average_linkage = cluster.AgglomerativeClustering(n_clusters=self.cluster_num,
affinity=self.AgglomerativeClustering['affinity'],
connectivity=self.AgglomerativeClustering['connectivity'],
compute_full_tree=self.AgglomerativeClustering['compute_full_tree'],
linkage=self.AgglomerativeClustering['linkage'],
pooling_func=self.AgglomerativeClustering['pooling_func'])
birch = cluster.Birch(threshold=self.Birch['threshold'], branching_factor=self.Birch['branching_factor'],
n_clusters=self.Birch['n_clusters'], compute_labels=self.Birch['compute_labels'],
copy=self.Birch['copy'])
clustering_algorithms = [two_means, spectral, ward,average_linkage,birch]
clustering_names = [
'MiniBatchKMeans',
'SpectralClustering', 'Ward', 'AgglomerativeClustering',
'Birch']
suanfa={}
for name, algorithm in zip(clustering_names, clustering_algorithms):
suanfa[name]={}
t0 = time.time()
algorithm.fit(dataset)
t1 = time.time()
suanfa[name]['time']=t1-t0
if hasattr(algorithm, 'labels_'):
y_pred = algorithm.labels_.astype(np.int)
#print(y_pred)
else:
y_pred = algorithm.predict(dataset)
#print(y_pred)
l=len(set(y_pred))
y_pred=np.array([y_pred[l] if y_pred[l]!=-1 else l-1 for l in range(len(y_pred))])
suanfa[name]['y_pred']=y_pred
if hasattr(algorithm, 'cluster_centers_'):
centers = algorithm.cluster_centers_
suanfa[name]['centroids']=centers
else:
#如果算法没有centers_属性，则采用聚类数求平均值方法。
p=[];c=np.zeros((self.cluster_num,self.dataset.shape[1]))
for j in range(self.cluster_num):
p=np.nonzero(suanfa[name]['y_pred']==j)[0]
if len(p)==0:
print("算法本身不允许聚类数达到如此之大，因为某些聚类中心没有样本。")
print("第%d个聚类中心没有聚类到样本点（聚类数从1开始）"%(j+1))
print("请重新选择聚类数")
data=np.mat(dataset)[p,:]
c[j]=np.mean(data,axis=0)
suanfa[name]['centroids']=c
key=list(suanfa.keys())
self.key=key
self.suanfa=suanfa
suanfa1= copy.deepcopy(suanfa) #suanfa1储存pca后的聚类中心
H={} #用于储存每个聚类的半径值(平均半径)
Max={} #最大距离
Min={}#最小距离
for j in range(5):
H[key[j]]=[]
Max[key[j]]=[]
Min[key[j]]=[]
if key[j]!='DBSCAN' and key[j]!='AffinityPropagation':
df=pd.concat([dataset,pd.DataFrame(suanfa[list(suanfa.keys())[j]]['centroids'])])
redu=PCA(n_components=2).fit_transform(df)
suanfa1[key[j]]['centroids']=redu[-self.cluster_num:,:]
#print(redu[-i:,:] )
#print(list(suanfa.keys())[j])
#print(df[:-self.cluster_num].shape, suanfa[keys[j]]['centroids'].shape,suanfa1[keys[j]]['centroids'].shape)
h=self.average(df[:-self.cluster_num],j,self.cluster_num) #计算平均距离，j是算法，i是聚类数
H[key[j]].append(h)
M1,M0=self.maxRminR1(df[:-self.cluster_num],j,self.cluster_num)
Max[key[j]].append(M1)
Min[key[j]].append(M0)
self.suanfa1=suanfa1
self.plot_Circle(self.cluster_num,H,Max,Min) #绘图函数，聚类数i
self.special_algorithm()
#self.Jun_level();
import random
data = load_iris()
dataset=pd.DataFrame(data['data'])
d=DrawCluster(dataset,3)
d.Drawnet()
x=len(dataset)
for i in [10,20,30,40,50,x]:
dataset=pd.DataFrame(data['data'])
dataindex=random.sample(range(x),i)
#print(dataindex,i)
dataset=dataset.iloc[dataindex,:]
#print(dataset)
d=DrawCluster(dataset,3)
d.Drawnet()