-
Notifications
You must be signed in to change notification settings - Fork 5
Expand file tree
/
Copy pathtime_icf.py
More file actions
259 lines (214 loc) · 10.3 KB
/
time_icf.py
File metadata and controls
259 lines (214 loc) · 10.3 KB
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
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.metrics import pairwise_distances
import copy
class ICF_topN:
def R_jaccard(self,I1,I2):
intersect = 1.0 * len(np.intersect1d(I1,I2))
if intersect == 0:
return 0
I1_g = 1.0 * len(I1) - intersect
I2_g = 1.0 * len(I2) - intersect
down = 1 + 1/intersect + I1_g/(1 + I1_g) + 1/(1 + I2_g)
return 1/down
def calculate_Rjaccard(self,data):
index_table = dict(zip(data.index.values,[0 for i in range(len(data.index))]))
for index,row in data.iterrows():
index_table[index] =row.dropna().index.values
rjaccard_table = pd.DataFrame(np.zeros((len(data.index),len(data.index)))+100,index=data.index,columns=data.index).to_dict()
for i in rjaccard_table:
for j in rjaccard_table:
if i == j:
rjaccard_table[i][j]=1
rjaccard_table[j][i]=1
continue
if rjaccard_table[i][j] != 100:
rjaccard_table[j][i] = rjaccard_table[i][j]
continue
rjaccard_table[j][i] = self.R_jaccard(index_table[j],index_table[i])
return rjaccard_table
def __init__(self,neighbor=10,similarity='pearson'):
self.neighbor = neighbor
self.similarity = similarity
def jaccard(self,n1,n2):
up = len(np.intersect1d(n1,n2))
down = len(np.union1d(n1,n2))
return up/down
def calculate_jaccard(self,data):
index_table = dict(zip(data.index.values,[0 for i in range(len(data.index))]))
for index,row in data.iterrows():
index_table[index] =row.dropna().index.values
jaccard_table = pd.DataFrame(index=data.index,columns=data.index).to_dict()
for i in data.index.values:
for j in data.index.values:
if i==j:
jaccard_table[i][j]=1
continue
if jaccard_table[i][j] >0:
jaccard_table[j][i] = jaccard_table[i][j]
continue
jaccard_table[j][i] = self.jaccard(index_table[j],index_table[i])
return jaccard_table
def calculate_similarity(self,data,IIF_t):
if self.similarity == 'pearson':
return IIF.to_dict()#pd.DataFrame(1- pairwise_distances(data.fillna(0),metric='correlation'),columns=data.index,index=data.index).to_dict()
if self.similarity == 'IIF':
# IIF = pd.read_csv('IIF_ta.csv',index_col=0)
# IIF = pd.DataFrame(IIF.values,index=IIF.index,columns=IIF.index)
return IIF_t.to_dict()
if self.similarity == 'jaccard':
# return self.calculate_jaccard(data)
return IIF_t.to_dict()
if self.similarity == 'rjaccard':
return self.calculate_Rjaccard(data)
def find_nearset_neighbor(self,movieId):
top_neighbor = sorted(self.similarity_set[movieId].items(), key=lambda e:e[1], reverse=True)[1:1+self.neighbor]
similar_index = [i[0] for i in top_neighbor]
return similar_index
def fit(self,data,IIF_t,time_table):
self.time_table = time_table
self.origin_data = data.copy()
self.dataset = data.fillna(0).to_dict()
self.similarity_set = self.calculate_similarity(data,IIF_t)
self.user_list = pd.DataFrame(index=item_based.columns,columns=['User_list']).to_dict()
for i in data.columns:
self.user_list['User_list'][i] = data[i].dropna().index.values
def recommend(self,userId,topN):
top = sorted(self.predict_set[userId].items(),key = lambda items:items[1],reverse=True)[:topN]
top_N = [i[0] for i in top]
return top_N
def calculate_Fscore(self,test_data,topN):
self.precision = []
self.recall = []
self.Fscore = []
self.coverage = 0
total_movie = set()
test_data = test_data[['userId','movieId','rating']]
for user in set(test_data.userId):
#top = sorted(self.predict_set[user].items(),key = lambda items:items[1],reverse=True)[:topN]
#top_N = [i[0] for i in top]
if user not in self.predict_set:
continue
top_N = self.recommend(user,topN)
total_movie.update(top_N)
test_set = test_data[(test_data['userId']==user) & (test_data['rating']>=3)]['movieId'].values#test_data.loc[test_data['userId']==user,'movieId'].values
if len(test_set)==0:
continue
inter = len(np.intersect1d(top_N,test_set))
precision = inter/topN
recall = inter/len(test_set)
if recall == 0:
fscore = 0
else:
fscore = (1+0.25)*(precision*recall)/(0.25*precision+recall)
self.precision.append(precision)
self.recall.append(recall)
self.Fscore.append(fscore)
self.coverage = len(total_movie)/650#len(self.similarity_set)
def predict_whole(self):
user_list = self.user_list
predict_set = copy.deepcopy(self.dataset)
for movie in self.dataset[list(self.dataset.keys())[0]]:
print(movie)
k_similar = self.find_nearset_neighbor(movie)
for user in self.dataset.keys():
if predict_set[user][movie] > 0:
predict_set[user][movie] = 0
continue
u_list = user_list['User_list'][user]
combine = np.intersect1d(u_list,k_similar)
p = 0
for k_index in combine:
if self.dataset[user][k_index] > 3:
p += self.similarity_set[movie][k_index]*1
predict_set[user][movie] = p
self.predict_set = predict_set
return self.predict_set
def predict_whole_time(self,current,b):
user_list = self.user_list
predict_set = copy.deepcopy(self.dataset)
for movie in self.dataset[list(self.dataset.keys())[0]]:
k_similar = self.find_nearset_neighbor(movie)
for user in self.dataset.keys():
if predict_set[user][movie] > 0:
predict_set[user][movie] = 0
continue
u_list = user_list['User_list'][user]
combine = np.intersect1d(u_list,k_similar)
p = 0
for k_index in combine:
if self.dataset[user][k_index] >= 3:
f = 1/(1 + b*abs(current - self.time_table[user][k_index]))
p += self.similarity_set[movie][k_index]*f
predict_set[user][movie] = p
self.predict_set = predict_set
return self.predict_set
def IIF(I1,I2,data,time_table,j,k,alpha):
intersect = np.intersect1d(I1,I2)
if len(intersect) == 0:
return 0
Nu = len(I1)
Nv = len(I2)
up = 0
for i in intersect:
f = 1/(1 + alpha*abs(time_table[i][k] -time_table[i][j]))
up += 1/(np.log(1+1.0*(len(data['User_list'][i])))) * f
return up/np.sqrt(1.0*Nu*Nv)
def IIF_table(movie_list,user_list,time_table,alpha):
data = user_list.to_dict()
a = dict(zip(list(movie_list['Movie_list'].keys()),[1000 for i in range(len(movie_list['Movie_list']))]))
table = dict(zip(list(movie_list['Movie_list'].keys()),[copy.deepcopy(a) for i in range(len(movie_list['Movie_list']))]))
for i in table:
# print(i)
for j in table:
if i == j:
table[j][i] = 1
continue
if table[i][j] != 1000:
table[j][i] = table[i][j]
continue
table[j][i] = IIF(movie_list['Movie_list'][j],movie_list['Movie_list'][i],data,time_table,j,i,alpha)
return pd.DataFrame(table,index=list(table.keys()),columns=list(table.keys()))
def jaccard(n1,n2,time_table,j,k,alpha):
inter = np.intersect1d(n1,n2)
if len(inter) == 0:
return 0
up = 0
for i in inter:
up+= 1*1/(1 + alpha*abs(time_table[i][k] - time_table[i][j]))
down = len(np.union1d(n1,n2))
return up/down
def calculate_jaccard(movie_list,user_list,time_table,alpha):
a = dict(zip(list(movie_list['Movie_list'].keys()),[1000 for i in range(len(movie_list['Movie_list']))]))
table = dict(zip(list(movie_list['Movie_list'].keys()),[copy.deepcopy(a) for i in range(len(movie_list['Movie_list']))]))
for i in table:
# print(i)
for j in table:
if i == j:
table[j][i] = 1
continue
if table[i][j] != 1000:
table[j][i] = table[i][j]
continue
table[j][i] = jaccard(movie_list['Movie_list'][j],movie_list['Movie_list'][i],time_table,j,i,alpha)
return pd.DataFrame(table,index=list(table.keys()),columns=list(table.keys()))
rating = pd.read_csv('./ml-1m/ratings.dat',sep='::',header=None,names=['userId','movieId','rating','timestamp'])
movie = pd.read_csv('./ml-1m/movies.dat',sep='::',header=None,names=['movieId','title','genres'])
rating = rating[(rating['userId']<3500) & (rating['movieId'] < 700)]
movie = movie[movie['movieId']<700]
from sklearn.model_selection import train_test_split
train,test = train_test_split(rating,test_size = 0.2,random_state=0)
item_based = train.pivot_table(index='movieId', columns='userId', values='rating')
# rating = rating.sort_values('timestamp')
time_table = train.pivot_table(index='movieId', columns='userId', values='timestamp')
movie_list = pd.DataFrame(index=item_based.index,columns=['Movie_list'])
for i in item_based.index.values:
print(i)
movie_list['Movie_list'][i] = item_based.loc[i].dropna().index.values
user_list = pd.DataFrame(index=item_based.columns,columns=['User_list'])
for i in item_based.columns:
user_list['User_list'][i] = item_based[i].dropna().index.values
IIF_t = IIF_table(movie_list,user_list,time_table,0.001)
jaccard_t = calculate_jaccard(movie_list,user_list,time_table,0.001)