-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathRecommendation System.py
189 lines (172 loc) · 7.59 KB
/
Recommendation System.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
from pyspark import SparkConf, SparkContext
import json
import sys
import time
import itertools
from collections import Counter
from itertools import chain
from collections import OrderedDict
import random
import collections
import math
from statistics import mean
from pyspark.mllib.recommendation import ALS, MatrixFactorizationModel, Rating
def fillmissing(pairs):
if pairs[1]==None:
pairs=(pairs[0],3.5)
return pairs
def predictcase2(userid,businessid):
if userid in umb_maps and businessid in bmu_maps:
otheritems_thisuser=dict(umb_maps[userid])
meannumber=mean(otheritems_thisuser.values())
related_users=bmu_maps[businessid]
nom=0
de=0
related_users=related_users[:40]
for i in related_users:
othersingleuser_items=dict(umb_maps[i[0]])
commen_items=set(othersingleuser_items.keys()) & set(otheritems_thisuser.keys())
commen_number=len(commen_items)
if commen_number<=1:
w_value=1
mean_two=i[1]
else:
nominater=0
denominater1=0
denominater2=0
one=0
two=0
for item in commen_items:
one=one+otheritems_thisuser[item]
two=two+othersingleuser_items[item]
mean_one=one/commen_number
mean_two=two/commen_number
for item in commen_items:
normalize_one=otheritems_thisuser[item]-mean_one
normalize_two=othersingleuser_items[item]-mean_two
nominater=nominater+normalize_one*normalize_two
denominater1=denominater1+normalize_one**2
denominater2=denominater2+normalize_two**2
denominater=math.sqrt(denominater1*denominater2)
if denominater==0:w_value=0
else: w_value=nominater/denominater
nom=nom+w_value*(i[1]-mean_two)
de=de+abs(w_value)
if de==0:predictvalue=meannumber
else:predictvalue=meannumber+(nom/de)
else: predictvalue=3.5
return (userid,businessid,predictvalue)
def predictcase3(userid,businessid):
if userid in umb_maps and businessid in bmu_maps:
timestamp1= time.time()
otherusers_thisitem=dict(bmu_maps[businessid])
related_businesses=umb_maps[userid]
nom=0
de=0
timestamp2= time.time()
related_businesses=related_businesses[:50]
for i in related_businesses:
othersingleitem_users=dict(bmu_maps[i[0]])
commen_user=set(othersingleitem_users.keys()) & set(otherusers_thisitem.keys())
commen_number=len(commen_user)
if commen_number<=1:
w_value=1
else:
nominater=0
denominater1=0
denominater2=0
one=0
two=0
for user in commen_user:
one=one+otherusers_thisitem[user]
two=two+othersingleitem_users[user]
mean_one=one/commen_number
mean_two=two/commen_number
for user in commen_user:
normalize_one=otherusers_thisitem[user]-mean_one
normalize_two=othersingleitem_users[user]-mean_two
nominater=nominater+normalize_one*normalize_two
denominater1=denominater1+normalize_one**2
denominater2=denominater2+normalize_two**2
denominater=math.sqrt(denominater1*denominater2)
if nominater<0:w_value=0
elif denominater==0:w_value=0
else: w_value=nominater/denominater
nom=nom+w_value*i[1]
de=de+abs(w_value)
timestamp3= time.time()
if de==0:predictvalue=3.5
else:predictvalue=nom/de
else: predictvalue=3.5
return (userid,businessid,predictvalue)
def task1(train_file_name,test_file_name,case_id,output_file_name):
conf = SparkConf().setMaster("local").setAppName("HW3")
sc=SparkContext(conf=conf)
startTime = time.time()
data=sc.textFile(train_file_name)
header = data.first()
data = data.filter(lambda row:row != header).map(lambda line: line.split(",")).persist()
val_data=sc.textFile(test_file_name)
val_header = val_data.first()
val_data = val_data.filter(lambda row:row != val_header).map(lambda line: line.split(",")).persist()
#dict
userid_dic_val = val_data.map(lambda x: x[0])
userid_dic = data.map(lambda x: x[0]).union(userid_dic_val).distinct().zipWithIndex().persist()
userid_dic_pos=userid_dic.collectAsMap()
userid_dic_neg=userid_dic.map(lambda x: (x[1],x[0])).collectAsMap()
businessid_dic_val = val_data.map(lambda x: x[1])
businessid_dic = t=data.map(lambda x: x[1]).union(businessid_dic_val).distinct().zipWithIndex().persist()#24731
businessid_dic_pos=businessid_dic.collectAsMap()
businessid_dic_neg=businessid_dic.map(lambda x: (x[1],x[0])).collectAsMap()
global umb_maps,bmu_maps
if int(case_id)==1:
ratings=data.map(lambda line: (userid_dic_pos[line[0]], businessid_dic_pos[line[1]],line[2]))\
.map(lambda l: Rating(int(l[0]), int(l[1]), float(l[2])))
val_ratings=val_data.map(lambda line: (userid_dic_pos[line[0]], businessid_dic_pos[line[1]],line[2]))\
.map(lambda l: Rating(int(l[0]), int(l[1]), float(l[2])))
rank = 20
numIterations = 20
model = ALS.train(ratings, rank, numIterations,lambda_=0.1)
valdata = val_ratings.map(lambda p: (p[0], p[1]))
predictions = model.predictAll(valdata).map(lambda r: ((r[0], r[1]), r[2]))
ratesAndPreds = val_ratings.map(lambda r: ((r[0], r[1]), r[2])).leftOuterJoin(predictions).map(lambda x: (x[0],fillmissing(x[1]))).persist()
output=ratesAndPreds.map(lambda x: (userid_dic_neg[x[0][0]],businessid_dic_neg[x[0][1]],x[1][1])).collect()
MSE=ratesAndPreds.map(lambda r: (r[1][0] - r[1][1])**2).mean()
RMSE=math.sqrt(MSE)
print("Root Mean Squared Error = " + str(RMSE))
elif int(case_id)==2:
ratings=data.map(lambda line: (userid_dic_pos[line[0]], businessid_dic_pos[line[1]],line[2])).distinct().persist()
umb_maps=ratings.map(lambda l: (int (l[0]), (int(l[1]), float(l[2])))).groupByKey().map(lambda e: (e[0],list(e[1]))).collectAsMap()
bmu_maps=ratings.map(lambda l: (int (l[1]), (int(l[0]), float(l[2])))).groupByKey().map(lambda e: (e[0],list(e[1]))).collectAsMap()
val_ratings=val_data.map(lambda line: (userid_dic_pos[line[0]], businessid_dic_pos[line[1]],line[2])).distinct()\
.map(lambda l: (int(l[0]), int(l[1]), float(l[2]))).persist()
valdata = val_ratings.map(lambda p: (predictcase2(p[0], p[1]),p[2])).persist()
output=valdata.map(lambda x: (userid_dic_neg[x[0][0]],businessid_dic_neg[x[0][1]],x[0][2])).collect()
MSE=valdata.map(lambda x:(x[0][2]-x[1])**2).mean()
RMSE=math.sqrt(MSE)
print("Root Mean Squared Error = " + str(RMSE))
elif int(case_id)==3:
ratings=data.map(lambda line: (userid_dic_pos[line[0]], businessid_dic_pos[line[1]],line[2])).distinct().persist()
umb_maps=ratings.map(lambda l: (int (l[0]), (int(l[1]), float(l[2])))).groupByKey().map(lambda e: (e[0],list(e[1]))).collectAsMap()
bmu_maps=ratings.map(lambda l: (int (l[1]), (int(l[0]), float(l[2])))).groupByKey().map(lambda e: (e[0],list(e[1]))).collectAsMap()
val_ratings=val_data.map(lambda line: (userid_dic_pos[line[0]], businessid_dic_pos[line[1]],line[2])).distinct()\
.map(lambda l: (int(l[0]), int(l[1]), float(l[2]))).persist()
valdata = val_ratings.map(lambda p: (predictcase3(p[0], p[1]),p[2])).persist()
output=valdata.map(lambda x: (userid_dic_neg[x[0][0]],businessid_dic_neg[x[0][1]],x[0][2])).collect()
MSE=valdata.map(lambda x:(x[0][2]-x[1])**2).mean()
RMSE=math.sqrt(MSE)
print("Root Mean Squared Error = " + str(RMSE))
else:
return print("This is an invaild case number, please enter 1~3")
with open(output_file_name, "w") as outfile1:
outfile1.write("user_id, business_id, prediction:\n")
for k in output:
i=','.join(str(w) for w in k)
outfile1.write(i+"\n")
endTime1 = time.time()
time1=endTime1-startTime
print("Duration " + str(time1))
if __name__ == "__main__":
task1(sys.argv[1], sys.argv[2],sys.argv[3],sys.argv[4])
#print(result)
#spark-submit xinyue_niu_task2.py yelp_train.csv yelp_val.csv 1 task2_case1.csv