-
Notifications
You must be signed in to change notification settings - Fork 0
/
icics.py
240 lines (175 loc) · 8.73 KB
/
icics.py
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
# -*- coding: utf-8 -*-
import numpy as np
import pandas as pd
from sklearn.cluster import KMeans
from sklearn.cross_validation import StratifiedKFold
import random
#compute euclid dist
def EuclidDis1(A,B):
'''
==============================================================
Description: A method to compute two matrix's euclid distance.
Usage：EuclidDis1(A,B)
==============================================================
Arguments:
A: A matrix
B: A matrix
the method return a np.matrix object.
'''
BT = B.transpose()
vecProd = A * BT
SqA = A.getA()**2
sumSqA = np.matrix(np.sum(SqA, axis=1))
sumSqAEx = np.tile(sumSqA.transpose(), (1, vecProd.shape[1]))
SqB = B.getA()**2
sumSqB = np.sum(SqB, axis=1)
sumSqBEx = np.tile(sumSqB, (vecProd.shape[0], 1))
SqED = sumSqBEx + sumSqAEx - 2*vecProd
ED = (SqED.getA())**0.5
return np.matrix(ED)
#model1
def recomm(dfold,df,bh,ywy,topn,ncluster,epsilon=0.001,init='k-means++',random_state=123,max_iter=1000,algorithm="auto"):
'''
=============================================================================================
Description: The method which can assign cases to a salesman.
Usage：recomm(dfold,df,bh,ywy,topn,ncluster,epsilon=0.001)
=============================================================================================
Arguments:
dfold: A DataFrame, the history case feature which this ywy has option.
df: A DataFrame, the new case feature which ready to assign.
bh: A list, the new case's id number which is one-to-one correspondence with df.
ywy: A string, the salesman's id number.
topn: An int number, how many cases should assign to the salesman.
ncluster: An int number, how many class should cluster.
epsilon: A little real number which avoid algorithmic errors
......
-----------------------------------------------------------------------------------------
more arguments find in the function of KMeans.
Require:
import numpy as np
import pandas as pd
from sklearn.cluster import KMeans
'''
clf = KMeans(n_clusters=ncluster,init = init,random_state=random_state,max_iter=max_iter,algorithm=algorithm).fit(dfold)
A = np.matrix(clf.cluster_centers_)
B = np.matrix(df)
dist = EuclidDis1(A, B)
dist = np.array(dist)
dist1 = dist.copy()
dist_sort = pd.DataFrame({'bh':bh,'cluster':list(np.argmax(dist1,axis=0)),'distance':list(np.max(dist1,axis=0))})
N = len(list(dist_sort.bh))
Ni = pd.DataFrame(dist_sort.cluster.value_counts())
Ni.reset_index(inplace=True)
Ni.rename(columns={"index":0,"cluster":1},inplace=True)
#Ni =pd.DataFrame(np.array(Ni0))
Ni.rename(columns={0:'class',1:'counts'},inplace=True)
dist_sort["group_sort"] = dist_sort['distance'].groupby(dist_sort['cluster']).rank(ascending=1,method='first')
dist_sort = dist_sort.merge(Ni,left_on='cluster',right_on='class',how="left")
dist_sort1 = dist_sort.groupby('cluster').apply(lambda x: x[x.group_sort <= topn*x.counts/(float(N)+epsilon)]) #可能会出bug
recommend = pd.DataFrame({'bh':np.array(dist_sort1['bh']),'ywy':list(np.unique(ywy))*len(np.array(dist_sort1['bh']))})
return recommend
#model2
def icic(dfold0,df0,topn0,bhywy0,ncluster,shuffle=0,epsilon=0.001,init='k-means++',random_state=123,max_iter=1000,algorithm="auto",path=0):
'''
==========================================================================
Description: The method which can assign cases to more than one salesmans.
Usage：icic(dfold0,df0,shuffle=0,topn0,bhywy0,ncluster,...)
==========================================================================
Arguments:
dfold0: A DataFrame, the history case feature which ywys has option.
dfold0 should incloud 'ywy0' and 'ajbh'and other features.
df0: A DataFrame, the new case feature which ready to assign.df0 should
incloud 'ajbh' and other features that the same as dfold0.
shuffle: 0 or 1, if 1,then random disorder of the order of df0 and bhywy0.
topn0: A DataFrame,inclouding two cols like: 'ywy0' and 'topn_ywy',topn_ywy means
how many cases should assign to this ywy0.
bhywy0: A DataFrame,inclouding two or one cols like: 'ywy0' and 'ajbh', and you can
not provide the 'ajbh' some times.
path: 0 or a string like '~/desktop/files.xlsx',it is a Excel file you want to save,
if path=0,means you do not need to save the assign result to you CP or disk.
......
------------------------------------------------------------------------------------------------
more arguments find in the function of recomm. You can use help(recomm) or ?(recomm) to get
support.
Require:
import numpy as np
import pandas as pd
from sklearn.cluster import KMeans
import random
'''
recommend_last =pd.DataFrame({"bh":[],"ywy":[]})
if shuffle==1:
nrowdf0 = range(df0.shape[0])
random.shuffle(nrowdf0)
df0 = df0.iloc[nrowdf0]
nrowbhywy0 = range(bhywy0.shape[0])
random.shuffle(nrowbhywy0)
bhywy0 = bhywy0.iloc[nrowbhywy0]
else:
pass
dflst = df0.copy()
ywylist = list(np.unique(list(bhywy0.ywy0)))
for ywy in ywylist:
if len(list(df0.ajbh)) >=1:
dfold = dfold0[dfold0.ywy0==ywy]
dfold1 = dfold.drop("ywy0",axis=1)
dfold2 = dfold1.drop("ajbh",axis=1)
bh = list(df0.ajbh)
df2 =df0.drop('ajbh',axis=1)
topn = int(np.unique(list(topn0[topn0.ywy0== ywy].topn_ywy)))
recommend = recomm(dfold2,df2,bh,ywy,topn,ncluster,epsilon=0.001,init='k-means++',random_state=123,max_iter=1000,algorithm="auto")
recommend_last = recommend_last.append(recommend)
df0 = df0[np.invert(df0['ajbh'].isin(list(recommend.bh)))]
recommend_last.reset_index(drop=True,inplace=True)
set1 = set(recommend_last.bh)
set2 = set(dflst.ajbh)
if len(set2-set1) == 0:
recommend_last1 = recommend_last
else:
dfno=pd.DataFrame({'bh':list(set2-set1),'ywy':list(np.random.choice ( ywylist,size=len(set2-set1)))})
frames = [recommend_last,dfno]
recommend_last1 = pd.concat(frames)
if path == 0:
pass
else:
writer = pd.ExcelWriter(path, engine='xlsxwriter')
recommend_last1.to_excel(writer, sheet_name='Sheet1')
writer.save()
return recommend_last1
#Training
def acc_mean(dfold0,bhywy0,topn0,ncluster,n_folds=5,shuffle=True,random_state=33,shuffle_icic=0,epsilon=0.001,init='k-means++',max_iter=1000,algorithm="auto",path=0):
'''
=============================================================================================
Description: The method which can training the icic model
Usage：acc_mean(labely,dfold0,bhywy0,ncluster,n_folds=5,shuffle=True,......)
=============================================================================================
Arguments:
labely: the training dataset label.
n_folds: the CV params.
shuffle: the CV params.
......
-----------------------------------------------------------------------------------------
more arguments find in the function of icic. You can use help(icic) or ?(icic) to get
support.
Require:
from sklearn.cross_validation import StratifiedKFold
'''
mean_acc=[]
labely = list(dfold0.ywy0)
skf = StratifiedKFold(labely,n_folds=n_folds,shuffle =shuffle,random_state=random_state)
for train_index,test_index in skf:
dfold1 = dfold0.iloc[train_index]
bhywy_test_real = dfold0.iloc[test_index][['ywy0','ajbh']]
df1 = dfold0.iloc[test_index].drop('ywy0',axis=1)
bhywy1 = bhywy0.iloc[train_index]
topn1 = topn0[topn0['ywy0'].isin (list(bhywy1.ywy0))]
validdata = icic(dfold1,df1,topn1,bhywy1,ncluster,shuffle=shuffle_icic,epsilon=epsilon,init=init,random_state=random_state,max_iter=max_iter,algorithm=algorithm,path=path)
predata = pd.merge(validdata,bhywy_test_real,left_on='bh',right_on='ajbh',how='left')
postive1 = 0
for i in np.arange(0,np.shape(predata)[0]):
if predata.ywy[i]==predata.ywy0[i]:
postive1+=1
acc = float(postive1)/len(test_index)
mean_acc.append(acc)
mean_acc1 = np.mean(mean_acc)
return mean_acc1