-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathpromoteTweets.py
154 lines (150 loc) · 7 KB
/
promoteTweets.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
#!/usr/bin/python
# tweets promotion baseline algorithm
# let us assume each user's retweet action is modeled by Gamma
# and in the situation of lacking data, we for now assume users and homogeneous.
# A simplified verions of patented MBT algorithm
# Copyright (c) 2013 All Right Reserved, Voxsup Inc.
import sys
import os
import string
import json
import numpy as np
import math
import random
# 1: promote, -1: not to promote, 0: it doesn't know what to do.
#@@@@@@ TO USERS@@@@@@@@
#
# $ python ./promoteTweets.py
# only need to modify SAMPLE, DATA,MCMCsteps,RESULTS
#
#
# assume all tweets from the same account and therefore their reach is assumeds to be equal.
# sample data contains multiple tweets. each tweet can have different number of datapoints.
"""
@@@@@@ sample output. the tweet with highest score is the best to promote. so sort the tweets in descending order
result prints the convergence in Gamma parameter space.
ONly the final values matter.
{'tweet3': 1, 'tweet2': 1, 'tweet1': 3, 'tweet4': 1}
{'tweet3': 11, 'tweet2': 11, 'tweet1': 33, 'tweet4': 11}
{'tweet3': 21, 'tweet2': 21, 'tweet1': 63, 'tweet4': 21}
{'tweet3': 31, 'tweet2': 23, 'tweet1': 81, 'tweet4': 31}
{'tweet3': 41, 'tweet2': 13, 'tweet1': 73, 'tweet4': 41}
{'tweet3': 51, 'tweet2': 3, 'tweet1': 63, 'tweet4': 51}
{'tweet3': 61, 'tweet2': -7, 'tweet1': 53, 'tweet4': 61}
{'tweet3': 71, 'tweet2': -17, 'tweet1': 43, 'tweet4': 71}
{'tweet3': 81, 'tweet2': -27, 'tweet1': 33, 'tweet4': 81}
{'tweet3': 91, 'tweet2': -37, 'tweet1': 23, 'tweet4': 91}
{'tweet3': 101, 'tweet2': -47, 'tweet1': -1, 'tweet4': 101}
{'tweet3': 111, 'tweet2': -57, 'tweet1': -31, 'tweet4': 111}
{'tweet3': 107, 'tweet2': -67, 'tweet1': -61, 'tweet4': 121}
{'tweet3': 97, 'tweet2': -77, 'tweet1': -91, 'tweet4': 131}
{'tweet3': 87, 'tweet2': -87, 'tweet1': -121, 'tweet4': 141}
{'tweet3': 77, 'tweet2': -97, 'tweet1': -151, 'tweet4': 151}
{'tweet3': 67, 'tweet2': -107, 'tweet1': -181, 'tweet4': 161}
{'tweet3': 57, 'tweet2': -117, 'tweet1': -211, 'tweet4': 171}
{'tweet3': 47, 'tweet2': -127, 'tweet1': -241, 'tweet4': 181}
{'tweet3': 37, 'tweet2': -137, 'tweet1': -271, 'tweet4': 191}
{'tweet3': 27, 'tweet2': -147, 'tweet1': -301, 'tweet4': 201}
{'tweet3': 17, 'tweet2': -157, 'tweet1': -331, 'tweet4': 211}
{'tweet3': 7, 'tweet2': -167, 'tweet1': -361, 'tweet4': 221}
{'tweet3': -3, 'tweet2': -177, 'tweet1': -391, 'tweet4': 231}
{'tweet3': -13, 'tweet2': -187, 'tweet1': -421, 'tweet4': 241}
{'tweet3': -23, 'tweet2': -197, 'tweet1': -451, 'tweet4': 251}
{'tweet3': -33, 'tweet2': -207, 'tweet1': -481, 'tweet4': 261}
{'tweet3': -43, 'tweet2': -217, 'tweet1': -511, 'tweet4': 271}
{'tweet3': -53, 'tweet2': -227, 'tweet1': -541, 'tweet4': 281}
{'tweet3': -63, 'tweet2': -237, 'tweet1': -571, 'tweet4': 291}
{'tweet3': -73, 'tweet2': -247, 'tweet1': -601, 'tweet4': 301}
{'tweet3': -83, 'tweet2': -257, 'tweet1': -631, 'tweet4': 311}
{'tweet3': -93, 'tweet2': -267, 'tweet1': -661, 'tweet4': 321}
{'tweet3': -103, 'tweet2': -277, 'tweet1': -691, 'tweet4': 331}
{'tweet3': -113, 'tweet2': -287, 'tweet1': -721, 'tweet4': 341}
{'tweet3': -123, 'tweet2': -297, 'tweet1': -751, 'tweet4': 341}
{'tweet3': -133, 'tweet2': -307, 'tweet1': -781, 'tweet4': 331}
{'tweet3': -143, 'tweet2': -317, 'tweet1': -811, 'tweet4': 321}
{'tweet3': -153, 'tweet2': -327, 'tweet1': -841, 'tweet4': 311}
{'tweet3': -163, 'tweet2': -337, 'tweet1': -871, 'tweet4': 301}
{'tweet3': -173, 'tweet2': -347, 'tweet1': -901, 'tweet4': 291}
{'tweet3': -183, 'tweet2': -357, 'tweet1': -931, 'tweet4': 281}
{'tweet3': -193, 'tweet2': -367, 'tweet1': -961, 'tweet4': 271}
{'tweet3': -203, 'tweet2': -377, 'tweet1': -991, 'tweet4': 261}
{'tweet3': -213, 'tweet2': -387, 'tweet1': -1021, 'tweet4': 251}
{'tweet3': -223, 'tweet2': -397, 'tweet1': -1051, 'tweet4': 241}
{'tweet3': -233, 'tweet2': -407, 'tweet1': -1081, 'tweet4': 231}
{'tweet3': -243, 'tweet2': -417, 'tweet1': -1111, 'tweet4': 221}
{'tweet3': -253, 'tweet2': -427, 'tweet1': -1141, 'tweet4': 211}
{'tweet3': -263, 'tweet2': -437, 'tweet1': -1171, 'tweet4': 201}
"""
SAMPLE= 1000 # REACH should be larger and any value_dimension values.
MCMCsteps = 500
DATA={'339824552404217856': [
# original input
{'vd1': 1, 'cd1': 520539},
{'vd1': 100, 'cd1': 521439},
{'vd1': 1, 'cd1': 522339},
{'vd1': 10000, 'cd1': 523239},
{'vd1': 1, 'cd1': 591639},
# dilute in time by minute
{'vd1': 1, 'cd1': 520539/60.0},
{'vd1': 100, 'cd1': 521439/60.0},
{'vd1': 1, 'cd1': 522339/60.0},
{'vd1': 10000, 'cd1': 523239/60.0},
{'vd1': 1, 'cd1': 591639/60.0},
# dilute in time by hour
{'vd1': 1, 'cd1': 520539/60.0/60.0},
{'vd1': 100, 'cd1': 521439/60.0/60.0},
{'vd1': 1, 'cd1': 522339/60.0/60.0},
{'vd1': 10000, 'cd1': 523239/60.0/60.0},
{'vd1': 1, 'cd1': 591639/60.0/60.0}
],
'341995162706259968': [
#original input
{'vd1': 0, 'cd1': 3025},
{'vd1': 0, 'cd1': 3925},
{'vd1': 0, 'cd1': 4825},
{'vd1': 0, 'cd1': 5725},
{'vd1': 0, 'cd1': 74125},
#dilute by minute
{'vd1': 0, 'cd1': 3025/60.0},
{'vd1': 0, 'cd1': 3925/60.0},
{'vd1': 0, 'cd1': 4825/60.0},
{'vd1': 0, 'cd1': 5725/60.0},
{'vd1': 0, 'cd1': 74125/60.0},
# dilute by hour
{'vd1': 0, 'cd1': 3025/60.0/60.0},
{'vd1': 0, 'cd1': 3925/60.0/60.0},
{'vd1': 0, 'cd1': 4825/60.0/60.0},
{'vd1': 0, 'cd1': 5725/60.0/60.0},
{'vd1': 0, 'cd1': 74125/60.0/60.0}
]
}
METRIC = {"cost_dimension_1":"time_elapsed_since_creation","value_dimension_1":"num_of_retweets"} # FYI
INIT_GAMMA={"shape":2.,"scale":2.} # initial Gamma distribution
RESULTS = {"339824552404217856":0,
"341995162706259968":0
}
# for tweet1: x1,...,x12 <= 10, 10< x13,...,x42<=20,
# 20 < x43,....x10000 < 200000
# all xi's are iid Gamma. The goal is to estimate Gamma's parameters
def drawGammaTest( shape, scale ):
s = np.random.gamma(shape, scale, SAMPLE)
for t in DATA.keys():
for row in DATA[t]:
cut_off = row['cd1']
# count the samples, whose response time is below cut off
res = sum([1 if val < cut_off else 0 for val in s ])
if res < row['vd1'] or SAMPLE < row['vd1']:
RESULTS[t] += 1 # being greater than means meets or exceeds expectation
else:
RESULTS[t] -= 1
### RESULTS = {"tweet1":1,"tweet2":1,.....}
#return sum(RESULTS) #
def MCMCitr( steps ):
for i in xrange(steps):
if (INIT_GAMMA['shape'] > 0 and INIT_GAMMA['scale'] > 0 ):
drawGammaTest(INIT_GAMMA['shape'],INIT_GAMMA['scale'])
INIT_GAMMA['shape'] += random.random() - 0.44
INIT_GAMMA['scale'] += random.random() - 0.44
if i%10 == 0:
print RESULTS
MCMCitr(MCMCsteps)