特征列 属性值 获取 vowpal wabbit 生成DNN 的训练测试数据
用户特征文件 userFeature.data 每 行 代 表 一 个 用 户 的 特 征 数 据， 格 式 为： “uid|features”，uid 和 features 用竖线“|”分隔。其中 feature 采用 vowpal wabbit（https://github.com/JohnLangford/vowpal_wabbit）格式： “feature_group1|feature_group2|feature_group3|...”。 每 个 feature_group 代表一个特征组，多个特征组之间也以竖线“|”分隔。一个特征组若包括多个值 则以空格分隔，格式为：“feature_group_name fea_name1 fea_name2 …”， 其中 fea_name 采用数据编号的格式。
特征列 属性值 获取
d = {}
with open(f, 'r') as fr:
for i in fr:
l = i.split('|')
for ii in l:
ll=ii.split(' ')
k=ll[0]
if k not in d:
d[k]=[]
for iii in ll[1:]:
iii_=int(iii)
if int(iii) not in d[k]:
d[k].append(iii_)
for k in d:
l=sorted(d[k],reverse=False)
print(k)
print(l)
批处理 减小运行时间
d, reduce_chk_counter, reduce_chk_step = {}, 0, 500000
with open(f, 'r') as fr:
for i in fr:
l = i.split('|')
for ii in l:
ll = ii.split(' ')
k = ll[0]
if k not in d:
d[k] = []
for iii in ll[1:]:
iii_ = int(iii)
d[k].append(iii_)
reduce_chk_counter += 1
if reduce_chk_counter == reduce_chk_step:
reduce_chk_counter = 0
for k in d:
d[k] = list(set(d[k]))
l = sorted(d[k], reverse=False)
print(k)
print(l)
for k in d:
d[k] = list(set(d[k]))
l = sorted(d[k], reverse=False)
print(k)
print(l)
res_f = 'toknowit.txt'
with open(res_f, 'w') as fw:
for k in d:
l = sorted(d[k], reverse=False)
print(k)
print(l)
s = '{}\n{}\n'.format(k, ','.join(l))
fw.write(s)
将数据冗余至内存，批处理，去重，释放内存
f = 'userFeature.data'
# 确定marriageStatus特征列的值
d, d1 = {}, {}
with open(f, 'r') as fr:
for i in fr:
break
l = i.split('|')
for ii in l:
if 'marriageStatus' in ii:
k = len(ii)
d[k] = ii
k = ii.split('marriageStatus')[-]
d1[k] = ii
for k in d:
print(k, d[k])
for k in d1:
print(k, d1[k]) '''
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus marriageStatus
''' d = {}
with open(f, 'r') as fr:
for i in fr:
break
l = i.split('|')
for ii in l:
ll = ii.split(' ')
k = ll[]
if k not in d:
d[k] = []
for iii in ll[:]:
iii_ = int(iii)
if int(iii) not in d[k]:
d[k].append(iii_) d, reduce_chk_counter, reduce_chk_step = {}, ,
with open(f, 'r') as fr:
for i in fr: break l = i.split('|') for ii in l:
ll = ii.split(' ')
k = ll[]
if k == 'uid':
continue
if k not in d:
d[k] = []
for iii in ll[:]:
iii_ = int(iii)
d[k].append(iii_) reduce_chk_counter +=
if reduce_chk_counter == reduce_chk_step:
reduce_chk_counter =
for k in d:
d[k] = list(set(d[k]))
l = sorted(d[k], reverse=False)
print(k)
print(l) for k in d:
break d[k] = list(set(d[k]))
l = sorted(d[k], reverse=False)
print(k)
print(l) res_f = 'toknowit.txt'
# with open(res_f, 'w') as fw:
# for k in d:
# l = sorted(d[k], reverse=False)
# print(k)
# print(l)
# s = '{}\n{}\n'.format(k, ','.join([str(i) for i in l]))
# fw.write(s)
cut_l = []
with open(res_f, 'r') as fr:
for i in fr:
l = i.replace('\n', '').split(',')[:]
cut_l.append(l) res_f_cut = 'toknowitCUT.txt' with open(res_f_cut, 'w') as fw:
for l in cut_l:
s = '{}\n'.format(','.join([str(i) for i in l]))
fw.write(s)
age
0,1,2,3,4,5
gender
0,1,2
marriageStatus
0,2,3,5,6,8,9,10,11,12,13,14,15
education
0,1,2,3,4,5,6,7
consumptionAbility
0,1,2
LBS
0,1,2,3,4,6,7,8,9,10,11,12,13,14,15,16,18,19,20,21,23,25,26,27,29,30,31,32,33,35,36,38,39,40,41,42,43,45,46,47,48,49,50,51,52,53,54,55,56,57,60,61,62,63,64,65,66,67,68,69,70,71,72,73,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,91,92,94,95,97,98,99,100,101,102,103,104,105,106,107,108,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,135,137,138,139,142,143,144,145,146,147,149,150,151,152,153,154,155,157,158,159,160,161,163,165,168,170,171,172,173,174,175,176,177,178,179,180,181,183,184,185,186,188,189,190,191,192,193,194,195,197,198,199,200,201,202,203,204,206,208,209,210,211,212,214,215,216,217,218,219,220,222,223,224,225,227,229,232,233,234,235,236
interest1
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
interest2
1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,78,79,80,81,82
interest5
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
kw1
2,3,8,13,17,19,21,28,29,39,41,42,43,46,56,59,65,68,69,70,71,72,74,86,87,88,90,92,95,100,101,105,106,109,111,112,113,119,121,123,125,131,133,136,139,141,142,143,145,150,152,156,157,162,163,166,169,172,173,174,176,177,180,181,183,184,185,186,191,199,203,204,209,211,214,216,230,235,240,242,243,246,249,260,263,265,268,269,271,272,278,279,283,284,289,291,292,295,302,303,304,307,313,317,321,322,323,331,336,341,343,344,351,354,357,358,359,366,367,369,370,372,373,375,376,377,378,380,381,382,390,391,393,396,401,402,406,407,408,409,411,414,417,423,429,433,434,437,438,441,442,443,449,456,464,465,468,472,473,475,477,478,480,482,485,486,487,495,496,497,504,506,507,511,513,521,522,526,532,536,541,542,546,547,560,561,563,566,567,575,576,578,581,584,588,592,594,604,605,610
kw2
2,6,7,9,10,11,12,14,21,22,23,25,26,30,34,38,40,41,42,43,44,46,47,50,55,56,62,63,66,69,70,71,72,74,75,76,77,78,80,81,84,85,87,89,90,91,94,95,100,112,114,116,117,118,119,121,123,124,127,128,129,130,133,135,137,142,143,144,148,149,151,153,154,156,157,158,163,168,171,174,176,177,180,183,184,186,192,193,195,196,197,200,202,203,215,216,217,219,221,223,228,229,235,237,238,240,241,246,248,250,255,258,260,263,266,269,272,275,276,278,280,286,287,290,294,295,296,297,301,302,303,305,313,317,321,323,327,330,333,334,338,339,340,341,343,344,345,347,354,358,359,363,366,368,369,371,374,375,377,378,380,383,384,386,391,393,394,395,398,399,400,403,404,405,408,409,412,413,417,418,422,427,433,436,438,440,442,445,447,448,451,453,454,455,456,457,459,461,462,463
topic1
0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,197,198,199
topic2
0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,197,198,199
ct
0,1,2,3,4
os
0,1,2
carrier
0,1,2,3
house
1
interest3
1,2,3,4,5,6,7,8,9,10
kw3
1,7,8,10,15,19,25,27,29,36,50,56,63,68,69,74,77,80,88,93,95,101,117,122,123,124,126,132,133,136,138,149,151,152,153,155,157,164,167,171,173,174,181,186,188,190,192,194,197,198,206,209,213,223,228,233,235,248,249,253,263,273,276,278,280,286,288,295,302,303,311,314,316,323,328,331,332,333,336,343,349,362,364,366,370,372,381,385,391,394,399,401,404,411,412,416,420,425,427,431,453,459,464,465,469,470,474,488,499,504,505,508,512,513,523,530,531,534,539,549,559,560,563,566,568,570,574,581,586,588,598,607,610,617,627,630,633,634,635,636,638,645,650,654,655,657,663,668,676,677,681,685,686,687,691,692,694,695,696,699,701,703,705,707,709,719,722,723,725,734,735,737,739,740,742,745,751,755,763,764,769,771,780,785,788,799,800,805,809,818,821,833,835,836,840,851,853,856,860,862
topic3
1,3,4,8,10,11,14,16,18,19,21,22,23,24,25,27,28,30,31,32,33,34,35,37,39,42,43,44,46,47,49,51,53,54,55,56,58,59,60,62,63,65,66,68,69,70,72,75,76,78,79,81,84,87,88,90,92,95,98,99,100,101,102,103,107,108,109,111,112,113,115,116,117,119,120,121,123,124,126,127,129,130,132,133,136,137,138,139,141,142,143,146,148,150,151,154,157,158,159,161,162,164,165,166,167,168,169,171,174,176,177,178,180,182,183,185,186,187,188,190,191,192,193,194,197,198,199,201,202,205,206,207,209,210,211,212,213,214,215,216,217,218,219,220,221,223,226,227,228,232,233,234,235,237,238,240,241,243,251,252,253,255,256,258,259,260,262,264,265,266,267,268,269,270,271,272,273,274,275,278,279,280,282,283,285,287,288,292,297,298,299,301,304,305,306,307,308,309,312,314
interest4
1,2,3,4,5,6,7,8,9,10
appIdAction
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
appIdInstall
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
res_f_cut = 'toknowitCUT.txt' # with open(res_f_cut, 'w') as fw:
# s = '\n'.join([','.join([str(i) for i in l]) for l in cut_l])
# fw.write(s) sparse_num_drop_max, is_odd_line = 50, True
with open(res_f_cut, 'r') as fr:
for i in fr:
l = i.replace('\n', '').split(',')
if is_odd_line == True:
is_odd_line = False
k = l[0]
else:
is_odd_line = True
if len(l) <= sparse_num_drop_max:
for ii in l:
s = '{}_{}'.format(k, ii)
print(s)
else:
print(k)
age_0
age_1
age_2
age_3
age_4
age_5
gender_0
gender_1
gender_2
marriageStatus_0
marriageStatus_2
marriageStatus_3
marriageStatus_5
marriageStatus_6
marriageStatus_8
marriageStatus_9
marriageStatus_10
marriageStatus_11
marriageStatus_12
marriageStatus_13
marriageStatus_14
marriageStatus_15
education_0
education_1
education_2
education_3
education_4
education_5
education_6
education_7
consumptionAbility_0
consumptionAbility_1
consumptionAbility_2
LBS
interest1
interest2
interest5
kw1
kw2
topic1
topic2
ct_0
ct_1
ct_2
ct_3
ct_4
os_0
os_1
os_2
carrier_0
carrier_1
carrier_2
carrier_3
house_1
interest3_1
interest3_2
interest3_3
interest3_4
interest3_5
interest3_6
interest3_7
interest3_8
interest3_9
interest3_10
kw3
topic3
interest4_1
interest4_2
interest4_3
interest4_4
interest4_5
interest4_6
interest4_7
interest4_8
interest4_9
interest4_10
appIdAction
appIdInstall
['uid', 'age', 'gender', 'marriageStatus', 'education', 'consumptionAbility', 'LBS', 'interest1', 'interest2', 'interest3', 'interest4', 'interest5', 'kw1', 'kw2', 'kw3', 'topic1', 'topic2', 'topic3', 'appIdInstall', 'appIdAction', 'ct', 'os', 'carrier', 'house'] ['age_0', 'age_1', 'age_2', 'age_3', 'age_4', 'age_5', 'gender_0', 'gender_1', 'gender_2', 'marriageStatus_0', 'marriageStatus_2', 'marriageStatus_3', 'marriageStatus_5', 'marriageStatus_6', 'marriageStatus_8', 'marriageStatus_9', 'marriageStatus_10', 'marriageStatus_11', 'marriageStatus_12', 'marriageStatus_13', 'marriageStatus_14', 'marriageStatus_15', 'education_0', 'education_1', 'education_2', 'education_3', 'education_4', 'education_5', 'education_6', 'education_7', 'consumptionAbility_0', 'consumptionAbility_1', 'consumptionAbility_2', 'LBS_0', 'interest1_0', 'interest2_0', 'interest3_1', 'interest3_2', 'interest3_3', 'interest3_4', 'interest3_5', 'interest3_6', 'interest3_7', 'interest3_8', 'interest3_9', 'interest3_10', 'interest4_1', 'interest4_2', 'interest4_3', 'interest4_4', 'interest4_5', 'interest4_6', 'interest4_7', 'interest4_8', 'interest4_9', 'interest4_10', 'interest5_0', 'kw1_0', 'kw2_0', 'kw3_0', 'topic1_0', 'topic2_0', 'topic3_0', 'appIdInstall_0', 'appIdAction_0', 'ct_0', 'ct_1', 'ct_2', 'ct_3', 'ct_4', 'os_0', 'os_1', 'os_2', 'carrier_0', 'carrier_1', 'carrier_2', 'carrier_3', 'house_1']
f = 'userFeature.data'
'''
17 marriageStatus 11
19 marriageStatus 2 13
20 marriageStatus 13 10
16 marriageStatus 0
21 marriageStatus 2 13 9
22 marriageStatus 12 13 9
23 marriageStatus 12 13 10 11 marriageStatus 11
5 13 marriageStatus 5 13
13 10 marriageStatus 13 10
10 marriageStatus 10
15 marriageStatus 15
0 marriageStatus 0
13 15 marriageStatus 13 15
12 13 marriageStatus 12 13
13 marriageStatus 13
6 13 marriageStatus 6 13
2 13 marriageStatus 2 13
13 9 marriageStatus 13 9
6 13 9 marriageStatus 6 13 9
2 13 9 marriageStatus 2 13 9
5 13 9 marriageStatus 5 13 9
12 13 9 marriageStatus 12 13 9
14 marriageStatus 14
12 13 10 marriageStatus 12 13 10
3 marriageStatus 3
15 10 marriageStatus 15 10
8 marriageStatus 8
6 13 10 marriageStatus 6 13 10
5 13 10 marriageStatus 5 13 10
13 10 9 marriageStatus 13 10 9
13 15 10 marriageStatus 13 15 10
2 13 10 marriageStatus 2 13 10 marriageStatus 0 2 5 6 8 9 10 11 12 13 15
''' def fw_s(f, s):
with open(f, 'w') as fw:
fw.write(s) # d 获取一个特征下有哪些值，如果值个数大于1，则考虑拆分该特征为各个子特征
# k_oreder_l 获取一级特征的顺序 res_f = 'toknowit.txt'
res_f_k_order = res_f.replace('.', 'KeyOrder.')
res_f_cut, children_val_max = res_f.replace('.', 'Cut.'), 50 to_write = True
to_write = False
if to_write:
d, reduce_chk_counter, reduce_chk_step, k_oreder_l = {}, 0, 500000, []
with open(f, 'r') as fr:
for i in fr:
l = i.replace('\n', '').split('|')
k_order_l_this = []
for ii in l:
ll = ii.split(' ')
k = ll[0]
k_order_l_this.append(k)
if k == 'uid':
continue
if k not in d:
d[k] = []
# order -->int
for iii in ll[1:]:
d[k].append(int(iii))
k_oreder_l.append(k_order_l_this)
reduce_chk_counter += 1
print(reduce_chk_counter)
if reduce_chk_counter % reduce_chk_step == 0:
# reduce_chk_counter = 0
for k in d:
d[k] = list(set(d[k]))
k_oreder_l = [e for i, e in enumerate(k_oreder_l) if
k_oreder_l.index(e) == i] # set() TypeError unhashable type:'list' for k in d:
d[k] = sorted(list(set(d[k])), reverse=False) k_oreder_l = [e for i, e in enumerate(k_oreder_l) if
k_oreder_l.index(e) == i] s = '\n'.join(['{}\n{}'.format(k, ','.join([str(i) for i in d[k]])) for k in d])
fw_s(res_f, s) s = '\n'.join(['{}\n{}'.format(k, ','.join([str(i) for i in d[k][0:children_val_max]])) for k in d])
fw_s(res_f_cut, s) s = '\n'.join(['|'.join(l) for l in k_oreder_l])
fw_s(res_f_k_order, s) with open(res_f_k_order, 'r') as fr:
ori_feature_l = [i.replace('\n', '').split('|') for i in fr]
feature_after_e_d = {}
for l in ori_feature_l:
for e in l:
if e not in feature_after_e_d:
feature_after_e_d[e] = []
feature_after_e_d[e] += l[l.index(e) + 1:]
feature_after_e_d[e] = list(set(feature_after_e_d[e])) feature_l = [k for k in sorted(feature_after_e_d, key=lambda e: len(feature_after_e_d[e]), reverse=True)]
print(feature_l) import re feature_reduce_l = [i if re.search('\d', i) is None else i[0:re.search('\d', i).endpos - 1] for i in feature_l]
# set 破坏了顺序
print(feature_reduce_l)
print(list(set(feature_reduce_l))) feature_reduce_l = [e for i, e in enumerate(feature_reduce_l) if feature_reduce_l.index(e) == i]
print(feature_reduce_l) sparse_num_drop_max, is_odd_line = 20, True
# 特征的属性值
res_d = {}
with open(res_f_cut, 'r') as fr:
for i in fr:
l = i.replace('\n', '').split(',')
if is_odd_line == True:
is_odd_line = False
k = l[0]
res_d[k] = []
else:
is_odd_line = True
if len(l) <= sparse_num_drop_max:
for ii in l:
res_d[k].append(ii)
else:
res_d[k].append(0)
feature_expand_l = [] feature_l.pop(feature_l.index('uid'))
for k in feature_l:
feature_expand_l += ['{}_{}'.format(k, i) for i in res_d[k]] print(feature_expand_l) dd = 5
import re, time f = 'userFeature.data' '''
17 marriageStatus 11
19 marriageStatus 2 13
20 marriageStatus 13 10
16 marriageStatus 0
21 marriageStatus 2 13 9
22 marriageStatus 12 13 9
23 marriageStatus 12 13 10 11 marriageStatus 11
5 13 marriageStatus 5 13
13 10 marriageStatus 13 10
10 marriageStatus 10
15 marriageStatus 15
0 marriageStatus 0
13 15 marriageStatus 13 15
12 13 marriageStatus 12 13
13 marriageStatus 13
6 13 marriageStatus 6 13
2 13 marriageStatus 2 13
13 9 marriageStatus 13 9
6 13 9 marriageStatus 6 13 9
2 13 9 marriageStatus 2 13 9
5 13 9 marriageStatus 5 13 9
12 13 9 marriageStatus 12 13 9
14 marriageStatus 14
12 13 10 marriageStatus 12 13 10
3 marriageStatus 3
15 10 marriageStatus 15 10
8 marriageStatus 8
6 13 10 marriageStatus 6 13 10
5 13 10 marriageStatus 5 13 10
13 10 9 marriageStatus 13 10 9
13 15 10 marriageStatus 13 15 10
2 13 10 marriageStatus 2 13 10 marriageStatus 0 2 5 6 8 9 10 11 12 13 15
''' def fw_s(f, s):
with open(f, 'w') as fw:
fw.write(s) # d 获取一个特征下有哪些值，如果值个数大于1，则考虑拆分该特征为各个子特征
# feature_order_l 获取一级特征的顺序 # 原始数据 特征挖掘
# 各个特征的子特征值,各个特征的子特征个数,各个特征的顺序,
f_feature = 'toknowit.txt'
f_feature_ele_num = f_feature.replace('.', 'EleNum.')
f_feature_incomplete, f_feature_complete = f_feature.replace('.', 'Incomplete.'), f_feature.replace('.',
'Complete.') # 原始数据 加工成生产数据
# 将一级特征的子特征升级为一级特征，但是限定原一级特征的子特征可以升级的特征数最大值；该值需结合算力、算法确定
f_reduce, f_output = f_feature.replace('.', 'Reduce.'), f_feature.replace('.', 'Output.') # <---以上生成的文件无参数，数据恒定 # 这个文件是第一阶段的结果文件，含参数文件，故加上时间戳 f_extend = f_feature.replace('.', 'Extend{}.'.format(int(time.time())))
to_write_immutable_file = True
# to_write_immutable_file = False
if to_write_immutable_file:
feature_d, feature_incomplete_rows_l, reduce_chk_counter, reduce_chk_step, = {}, [], 0, 300000 # 打开GB文件，得到2个集合：获取不完整特征序列的集合、特征-特征值的集合 with open(f, 'r') as fr:
for i in fr:
l = i.replace('\n', '').split('|')
feature_incomplete_rows_l_this = []
for ii in l:
ll = ii.split(' ')
k = ll[0]
feature_incomplete_rows_l_this.append(k)
if k == 'uid':
continue
if k not in feature_d:
feature_d[k] = []
# order -->int
for iii in ll[1:]:
feature_d[k].append(int(iii))
feature_incomplete_rows_l.append(feature_incomplete_rows_l_this)
reduce_chk_counter += 1
print(reduce_chk_counter)
if reduce_chk_counter % reduce_chk_step == 0:
# reduce_chk_counter = 0 #从节约内存的角度，应重置为0，测试阶段观察分析进度和数据总数
for k in feature_d:
feature_d[k] = list(set(feature_d[k]))
feature_incomplete_rows_l = [e for i, e in enumerate(feature_incomplete_rows_l) if
feature_incomplete_rows_l.index(e) == i]
# subset TEST
break for k in feature_d:
feature_d[k] = sorted(list(set(feature_d[k])), reverse=False) feature_incomplete_rows_l = [e for i, e in enumerate(feature_incomplete_rows_l) if
feature_incomplete_rows_l.index(e) == i] s = '\n'.join([','.join(l) for l in feature_incomplete_rows_l])
fw_s(f_feature_incomplete, s) feature_after_e_d = {}
for l in feature_incomplete_rows_l:
for e in l:
if e not in feature_after_e_d:
feature_after_e_d[e] = []
feature_after_e_d[e] += l[l.index(e) + 1:]
feature_after_e_d[e] = list(set(feature_after_e_d[e])) # 原始一级特征b
feature_complete_l = [k for k in
sorted(feature_after_e_d, key=lambda e: len(feature_after_e_d[e]), reverse=True)]
print(feature_complete_l) s = '\n'.join(feature_complete_l)
fw_s(f_feature_complete, s) print(feature_complete_l) feature_d_ = {}
for feature in feature_complete_l:
if feature == 'uid':
continue
feature_d_[feature] = feature_d[feature]
del feature_d
feature_d = feature_d_ s = '\n'.join(['{}\n{}'.format(k, ','.join([str(i) for i in feature_d[k]])) for k in feature_d])
fw_s(f_feature, s) s = '\n'.join(['{}\n{}'.format(k, len(feature_d[k])) for k in feature_d])
fw_s(f_feature_ele_num, s) # 原始数据持久化完毕<--- # --->对原始数据做加工，生成新数据 # 0级别特征,将原始一级特征中"kw1,kw2,"合并
feature_reduce_l = [i if re.search('\d', i) is None else i[0:re.search('\d', i).endpos - 1] for i in
feature_complete_l]
# set 破坏了顺序
print(feature_reduce_l)
print(list(set(feature_reduce_l))) feature_reduce_l = [e for i, e in enumerate(feature_reduce_l) if feature_reduce_l.index(e) == i]
print(feature_reduce_l)
s = '\n'.join(feature_reduce_l)
fw_s(f_reduce, s) relative_, absolute_ = 2 / 3, 50
sparse_num_drop_max = min(
[absolute_,
max(sorted([len(feature_d[k]) for k in feature_d], reverse=False)[0:int(len(feature_d) * relative_)])]) s = '\n'.join(
['{}\n{}'.format(k, ','.join([str(i) for i in feature_d[k][0:sparse_num_drop_max]])) for k in feature_d]) fw_s(f_output, s) # 特征的属性值
feature_extend_d = {}
is_odd_line = True
with open(f_output, 'r') as fr:
for i in fr:
l = i.replace('\n', '').split(',')
if is_odd_line == True:
is_odd_line = False
k = l[0]
feature_extend_d[k] = []
else:
is_odd_line = True
if len(l) <= sparse_num_drop_max:
for ii in l:
feature_extend_d[k].append(ii)
else:
feature_extend_d[k].append(0)
feature_extend_l = [] feature_complete_l.pop(feature_complete_l.index('uid')) feature_extend_l = '|'.join(
['|'.join(['{}_{}'.format(k, str(i)) for i in feature_extend_d[k]]) for k in feature_extend_d]).split('|') print(feature_extend_l)
s = ','.join(feature_extend_l)
fw_s(f_extend, s) # 生成缩小的数据集，测试spark join 广告特征做广播 # feature_extend_l, f_extend = [], 'toknowitExtend1526836898.txt'
#
# with open(f_extend, 'r') as fr:
# feature_extend_l = [i.replace('\n', '') for i in fr][0].split(',') d = 8
age_0,age_1,age_2,age_3,age_4,age_5,gender_0,gender_1,gender_2,marriageStatus_0,marriageStatus_2,marriageStatus_3,marriageStatus_5,marriageStatus_6,marriageStatus_9,marriageStatus_10,marriageStatus_11,marriageStatus_12,marriageStatus_13,marriageStatus_14,marriageStatus_15,education_0,education_1,education_2,education_3,education_4,education_5,education_6,education_7,consumptionAbility_0,consumptionAbility_1,consumptionAbility_2,LBS_0,LBS_1,LBS_4,LBS_6,LBS_7,LBS_8,LBS_9,LBS_13,LBS_14,LBS_15,LBS_16,LBS_18,LBS_19,LBS_21,LBS_25,LBS_27,LBS_29,LBS_32,LBS_33,LBS_35,LBS_38,LBS_39,LBS_41,LBS_43,LBS_45,LBS_46,LBS_47,LBS_48,LBS_49,LBS_52,LBS_54,LBS_56,LBS_57,LBS_61,LBS_62,LBS_63,LBS_64,LBS_66,LBS_69,LBS_71,LBS_72,LBS_73,LBS_75,LBS_77,LBS_78,LBS_81,LBS_83,LBS_84,LBS_85,LBS_86,interest1_1,interest1_2,interest1_3,interest1_4,interest1_5,interest1_6,interest1_7,interest1_8,interest1_9,interest1_10,interest1_11,interest1_12,interest1_13,interest1_14,interest1_15,interest1_16,interest1_17,interest1_18,interest1_19,interest1_20,interest1_21,interest1_22,interest1_23,interest1_24,interest1_25,interest1_26,interest1_27,interest1_28,interest1_29,interest1_30,interest1_31,interest1_32,interest1_33,interest1_34,interest1_35,interest1_36,interest1_37,interest1_38,interest1_39,interest1_40,interest1_41,interest1_42,interest1_43,interest1_44,interest1_45,interest1_46,interest1_47,interest1_48,interest1_49,interest1_50,interest2_1,interest2_2,interest2_3,interest2_4,interest2_5,interest2_6,interest2_7,interest2_8,interest2_9,interest2_10,interest2_11,interest2_12,interest2_13,interest2_14,interest2_15,interest2_16,interest2_17,interest2_18,interest2_19,interest2_20,interest2_21,interest2_22,interest2_23,interest2_24,interest2_25,interest2_26,interest2_27,interest2_28,interest2_29,interest2_30,interest2_31,interest2_32,interest2_33,interest2_35,interest2_36,interest2_37,interest2_38,interest2_39,interest2_40,interest2_41,interest2_42,interest2_43,interest2_44,interest2_45,interest2_46,interest2_47,interest2_48,interest2_49,interest2_50,interest2_51,interest3_1,interest3_2,interest3_3,interest3_4,interest3_5,interest3_6,interest3_7,interest3_8,interest3_9,interest3_10,interest4_1,interest4_2,interest4_3,interest4_4,interest4_5,interest4_6,interest4_7,interest4_8,interest4_9,interest4_10,interest5_1,interest5_2,interest5_3,interest5_4,interest5_5,interest5_6,interest5_7,interest5_8,interest5_9,interest5_10,interest5_11,interest5_12,interest5_13,interest5_14,interest5_15,interest5_16,interest5_17,interest5_18,interest5_19,interest5_20,interest5_21,interest5_22,interest5_23,interest5_24,interest5_25,interest5_26,interest5_27,interest5_28,interest5_29,interest5_30,interest5_31,interest5_32,interest5_33,interest5_34,interest5_35,interest5_36,interest5_37,interest5_38,interest5_39,interest5_40,interest5_41,interest5_42,interest5_43,interest5_44,interest5_45,interest5_46,interest5_47,interest5_48,interest5_49,interest5_50,kw1_13,kw1_19,kw1_28,kw1_69,kw1_70,kw1_72,kw1_87,kw1_92,kw1_105,kw1_106,kw1_109,kw1_119,kw1_121,kw1_123,kw1_133,kw1_136,kw1_145,kw1_152,kw1_157,kw1_163,kw1_169,kw1_176,kw1_177,kw1_180,kw1_181,kw1_191,kw1_209,kw1_235,kw1_242,kw1_249,kw1_278,kw1_279,kw1_289,kw1_295,kw1_313,kw1_317,kw1_321,kw1_336,kw1_341,kw1_344,kw1_354,kw1_358,kw1_366,kw1_367,kw1_370,kw1_376,kw1_378,kw1_380,kw1_382,kw1_391,kw2_2,kw2_10,kw2_11,kw2_34,kw2_46,kw2_47,kw2_50,kw2_55,kw2_62,kw2_63,kw2_69,kw2_70,kw2_76,kw2_87,kw2_91,kw2_100,kw2_114,kw2_116,kw2_117,kw2_123,kw2_124,kw2_127,kw2_129,kw2_135,kw2_137,kw2_142,kw2_144,kw2_151,kw2_158,kw2_163,kw2_168,kw2_174,kw2_177,kw2_180,kw2_184,kw2_192,kw2_196,kw2_197,kw2_200,kw2_202,kw2_215,kw2_216,kw2_217,kw2_223,kw2_235,kw2_237,kw2_240,kw2_241,kw2_246,kw2_250,kw3_7,kw3_27,kw3_29,kw3_68,kw3_80,kw3_88,kw3_95,kw3_101,kw3_138,kw3_171,kw3_186,kw3_197,kw3_198,kw3_206,kw3_213,kw3_223,kw3_248,kw3_263,kw3_273,kw3_302,kw3_316,kw3_336,kw3_349,kw3_362,kw3_381,kw3_401,kw3_412,kw3_416,kw3_453,kw3_465,kw3_470,kw3_488,kw3_513,kw3_534,kw3_549,kw3_560,kw3_570,kw3_581,kw3_586,kw3_598,kw3_610,kw3_627,kw3_633,kw3_638,kw3_668,kw3_685,kw3_692,kw3_694,kw3_695,kw3_701,topic1_0,topic1_1,topic1_2,topic1_3,topic1_4,topic1_5,topic1_6,topic1_7,topic1_9,topic1_10,topic1_11,topic1_12,topic1_13,topic1_14,topic1_15,topic1_16,topic1_17,topic1_18,topic1_19,topic1_20,topic1_21,topic1_22,topic1_23,topic1_24,topic1_25,topic1_26,topic1_27,topic1_28,topic1_29,topic1_30,topic1_31,topic1_32,topic1_33,topic1_34,topic1_35,topic1_36,topic1_37,topic1_38,topic1_39,topic1_40,topic1_41,topic1_42,topic1_43,topic1_44,topic1_45,topic1_46,topic1_47,topic1_48,topic1_49,topic1_50,topic2_0,topic2_2,topic2_3,topic2_4,topic2_5,topic2_6,topic2_7,topic2_9,topic2_10,topic2_11,topic2_13,topic2_14,topic2_15,topic2_16,topic2_17,topic2_19,topic2_20,topic2_21,topic2_22,topic2_24,topic2_25,topic2_26,topic2_27,topic2_28,topic2_29,topic2_30,topic2_31,topic2_32,topic2_33,topic2_34,topic2_35,topic2_36,topic2_39,topic2_40,topic2_41,topic2_42,topic2_43,topic2_44,topic2_45,topic2_46,topic2_47,topic2_48,topic2_49,topic2_50,topic2_51,topic2_52,topic2_53,topic2_54,topic2_55,topic2_56,topic3_3,topic3_10,topic3_11,topic3_14,topic3_18,topic3_24,topic3_28,topic3_30,topic3_31,topic3_33,topic3_39,topic3_42,topic3_43,topic3_47,topic3_53,topic3_55,topic3_56,topic3_58,topic3_59,topic3_60,topic3_62,topic3_66,topic3_68,topic3_70,topic3_72,topic3_76,topic3_78,topic3_79,topic3_81,topic3_84,topic3_87,topic3_90,topic3_92,topic3_99,topic3_100,topic3_101,topic3_109,topic3_111,topic3_112,topic3_119,topic3_121,topic3_123,topic3_124,topic3_127,topic3_130,topic3_136,topic3_137,topic3_138,topic3_139,topic3_141,appIdInstall_1,appIdInstall_4,appIdInstall_6,appIdInstall_9,appIdInstall_10,appIdInstall_11,appIdInstall_12,appIdInstall_15,appIdInstall_16,appIdInstall_17,appIdInstall_19,appIdInstall_21,appIdInstall_23,appIdInstall_26,appIdInstall_27,appIdInstall_28,appIdInstall_29,appIdInstall_32,appIdInstall_34,appIdInstall_35,appIdInstall_39,appIdInstall_40,appIdInstall_41,appIdInstall_42,appIdInstall_43,appIdInstall_44,appIdInstall_45,appIdInstall_47,appIdInstall_48,appIdInstall_49,appIdInstall_51,appIdInstall_52,appIdInstall_55,appIdInstall_56,appIdInstall_57,appIdInstall_58,appIdInstall_60,appIdInstall_61,appIdInstall_62,appIdInstall_63,appIdInstall_65,appIdInstall_67,appIdInstall_68,appIdInstall_69,appIdInstall_70,appIdInstall_71,appIdInstall_73,appIdInstall_74,appIdInstall_76,appIdInstall_77,appIdAction_2,appIdAction_4,appIdAction_5,appIdAction_7,appIdAction_8,appIdAction_11,appIdAction_13,appIdAction_14,appIdAction_16,appIdAction_17,appIdAction_27,appIdAction_30,appIdAction_32,appIdAction_33,appIdAction_34,appIdAction_35,appIdAction_36,appIdAction_37,appIdAction_38,appIdAction_39,appIdAction_40,appIdAction_41,appIdAction_43,appIdAction_44,appIdAction_45,appIdAction_47,appIdAction_50,appIdAction_51,appIdAction_52,appIdAction_53,appIdAction_55,appIdAction_56,appIdAction_60,appIdAction_62,appIdAction_65,appIdAction_66,appIdAction_69,appIdAction_70,appIdAction_71,appIdAction_72,appIdAction_74,appIdAction_75,appIdAction_76,appIdAction_77,appIdAction_80,appIdAction_81,appIdAction_83,appIdAction_84,appIdAction_85,appIdAction_91,ct_0,ct_1,ct_2,ct_3,ct_4,os_0,os_1,os_2,carrier_0,carrier_1,carrier_2,carrier_3,house_1
JOIN 操作转移至Spark
import re, time f = 'userFeature.data' '''
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus
marriageStatus marriageStatus
''' def fw_s(f, s):
with open(f, 'w') as fw:
fw.write(s) # d 获取一个特征下有哪些值，如果值个数大于1，则考虑拆分该特征为各个子特征
# feature_order_l 获取一级特征的顺序 # 原始数据 特征挖掘
# 各个特征的子特征值,各个特征的子特征个数,各个特征的顺序,
f_feature = 'toknowit.txt'
f_feature_ele_num = f_feature.replace('.', 'EleNum.')
f_feature_incomplete, f_feature_complete = f_feature.replace('.', 'Incomplete.'), f_feature.replace('.',
'Complete.') # 原始数据 加工成生产数据
# 将一级特征的子特征升级为一级特征，但是限定原一级特征的子特征可以升级的特征数最大值；该值需结合算力、算法确定
f_reduce, f_output = f_feature.replace('.', 'Reduce.'), f_feature.replace('.', 'Output.') # <---以上生成的文件无参数，数据恒定 # 这个文件是第一阶段的结果文件，含参数文件，故加上时间戳 f_extend = f_feature.replace('.', 'Extend{}.'.format(int(time.time())))
to_write_immutable_file = True
# to_write_immutable_file = False
if to_write_immutable_file:
feature_d, feature_incomplete_rows_l, reduce_chk_counter, reduce_chk_step, = {}, [], , # 打开GB文件，得到2个集合：获取不完整特征序列的集合、特征-特征值的集合 with open(f, 'r') as fr:
for i in fr:
l = i.replace('\n', '').split('|')
feature_incomplete_rows_l_this = []
for ii in l:
ll = ii.split(' ')
k = ll[]
feature_incomplete_rows_l_this.append(k)
if k == 'uid':
continue
if k not in feature_d:
feature_d[k] = []
# order -->int
for iii in ll[:]:
feature_d[k].append(int(iii))
feature_incomplete_rows_l.append(feature_incomplete_rows_l_this)
reduce_chk_counter +=
print(reduce_chk_counter)
if reduce_chk_counter % reduce_chk_step == :
# reduce_chk_counter = #从节约内存的角度，应重置为0，测试阶段观察分析进度和数据总数
for k in feature_d:
feature_d[k] = list(set(feature_d[k]))
feature_incomplete_rows_l = [e for i, e in enumerate(feature_incomplete_rows_l) if
feature_incomplete_rows_l.index(e) == i]
# subset TEST
# break for k in feature_d:
feature_d[k] = sorted(list(set(feature_d[k])), reverse=False) feature_incomplete_rows_l = [e for i, e in enumerate(feature_incomplete_rows_l) if
feature_incomplete_rows_l.index(e) == i] s = '\n'.join([','.join(l) for l in feature_incomplete_rows_l])
fw_s(f_feature_incomplete, s) feature_after_e_d = {}
for l in feature_incomplete_rows_l:
for e in l:
if e not in feature_after_e_d:
feature_after_e_d[e] = []
feature_after_e_d[e] += l[l.index(e) + :]
feature_after_e_d[e] = list(set(feature_after_e_d[e])) # 原始一级特征b
feature_complete_l = [k for k in
sorted(feature_after_e_d, key=lambda e: len(feature_after_e_d[e]), reverse=True)]
print(feature_complete_l) s = '\n'.join(feature_complete_l)
fw_s(f_feature_complete, s) print(feature_complete_l) feature_d_ = {}
for feature in feature_complete_l:
if feature == 'uid':
continue
feature_d_[feature] = feature_d[feature]
del feature_d
feature_d = feature_d_ s = '\n'.join(['{}\n{}'.format(k, ','.join([str(i) for i in feature_d[k]])) for k in feature_d])
fw_s(f_feature, s) s = '\n'.join(['{}\n{}'.format(k, len(feature_d[k])) for k in feature_d])
fw_s(f_feature_ele_num, s) # 原始数据持久化完毕<--- # --->对原始数据做加工，生成新数据 # 0级别特征,将原始一级特征中"kw1,kw2,"合并
feature_reduce_l = [i if re.search('\d', i) is None else i[:re.search('\d', i).endpos - ] for i in
feature_complete_l]
# set 破坏了顺序
print(feature_reduce_l)
print(list(set(feature_reduce_l))) feature_reduce_l = [e for i, e in enumerate(feature_reduce_l) if feature_reduce_l.index(e) == i]
print(feature_reduce_l)
s = '\n'.join(feature_reduce_l)
fw_s(f_reduce, s) relative_, absolute_ = / ,
sparse_num_drop_max = min(
[absolute_,
max(sorted([len(feature_d[k]) for k in feature_d], reverse=False)[:int(len(feature_d) * relative_)])]) s = '\n'.join(
['{}\n{}'.format(k, ','.join([str(i) for i in feature_d[k][:sparse_num_drop_max]])) for k in feature_d]) fw_s(f_output, s) # 特征的属性值
feature_extend_d = {}
is_odd_line = True
with open(f_output, 'r') as fr:
for i in fr:
l = i.replace('\n', '').split(',')
if is_odd_line == True:
is_odd_line = False
k = l[]
feature_extend_d[k] = []
else:
is_odd_line = True
if len(l) <= sparse_num_drop_max:
for ii in l:
feature_extend_d[k].append(ii)
else:
feature_extend_d[k].append()
feature_extend_l = [] feature_complete_l.pop(feature_complete_l.index('uid')) feature_extend_l = '|'.join(
['|'.join(['{}_{}'.format(k, str(i)) for i in feature_extend_d[k]]) for k in feature_extend_d]).split('|') print(feature_extend_l)
s = ','.join(feature_extend_l)
fw_s(f_extend, s) # 生成缩小的数据集，测试spark join 广告特征做广播
ori_l, extend_l = [], []
with open('toknowitComplete.txt', 'r')as fr:
ori_l = [i.replace('\n', '') for i in fr] feature_extend_l, f_feature_extend = [], f_extend
with open(f_feature_extend, 'r')as fr:
extend_l = [i.replace('\n', '').split(',') for i in fr][] ori_extend_d = {} for ori in ori_l:
for extend_ in extend_l:
if ori in extend_:
if ori not in ori_extend_d:
ori_extend_d[ori] = {}
extend_d = {extend_: }
ori_extend_d[ori][extend_] =
import copy ori_extend_d_ = copy.deepcopy(ori_extend_d)
for i in ori_extend_d_['age']:
if 'marriageStatus' in i:
del ori_extend_d['age'][i]
del ori_extend_d_ '''
-生成数据元结构，末端值全为0 -拿到每行数据，去更新末端值 '''
c_ =
rows_d_l = []
with open(f, 'r') as fr:
for i in fr:
# c_ +=
# if c_ == :
# break ori_row_l = i.replace('\n', '').split('|')
ori_extend_d_this = copy.deepcopy(ori_extend_d)
uid_d = {}
for ii in ori_row_l:
l = ii.split(' ')
print(l)
feature_ori, val_l = l[], l[:]
if feature_ori == 'uid':
uid = val_l[]
continue
if len(ori_extend_d[feature_ori]) == :
for feature_sub in ori_extend_d_this[feature_ori]:
print(feature_sub)
ori_extend_d_this[feature_ori][feature_sub] = if int(val_l[]) > else
else:
for val_ in val_l:
feature_sub = '{}_{}'.format(feature_ori, val_)
print(feature_sub)
if feature_sub in ori_extend_d_this[feature_ori]: ###多余的校验
ori_extend_d_this[feature_ori][feature_sub] = # ????
uid_d[uid] = ori_extend_d_this
del ori_extend_d_this
rows_d_l.append(uid_d)
del uid_d s_l = []
f_userdata_extend = f.replace('.data', '{}.data'.format(int(time.time())))
for d in rows_d_l:
for uid in d:
c_ +=
l = []
d_ = d[uid]
for feature_ in d_:
for feature_sub in d_[feature_]:
l.append(d_[feature_][feature_sub])
s = '{},{}'.format(uid, ','.join([str(i) for i in l]))
s_l.append(s)
fw_s(f_userdata_extend, '\n'.join(s_l))
print(c_) '''
gen JOIN data FOR DNN '''
f_user = 'userFeature.data'
f_ad = 'adFeature.csv'
f_user_extend = f_userdata_extend
f_train = 'train.csv'
f_test = 'test2.csv' '''
gen head
'''
csv_head = 'advertiserId,campaignId,creativeId,creativeSize,adCategoryId,productId,productType'
f_toknowitExtend = 'toknowitExtend1527038949.txt'
try:
with open(f_toknowitExtend, 'r') as fr:
for i in fr:
csv_head = 'label,{},{}'.format(i.replace('\n', ''), csv_head)
print(csv_head)
except Exception as e:
print(e)
csv_head = '' # no file
'''
get dict
''' ad_d = {}
with open(f_ad, 'r') as fr:
for i in fr:
if 'aid' in i:
continue
l = i.replace('\n', '').split(',')
aid = l[]
ad_d[aid] = ','.join(l[:]) uid_d = {}
with open(f_user_extend, 'r') as fr:
for i in fr:
if 'aid' in i:
continue
l = i.replace('\n', '').split(',')
uid = l[]
uid_d[uid] = ','.join(l[:])
'''
gen train data
'''
dnn_csvTRAIN = 'dnn_csvTRAIN{}.csv'.format(int(time.time()))
with open(dnn_csvTRAIN, 'w') as fa:
fa.write(csv_head)
with open(f_train, 'r') as fr:
for i in fr:
if 'aid' in i:
continue
try:
l = i.replace('\n', '').replace(' ', '').split(',')
print(l)
aid, uid, label = l
s = '{},{},{}\n'.format(label, uid_d[uid], ad_d[aid])
with open(dnn_csvTRAIN, 'a') as fa:
fa.write(s)
except Exception as e:
print(e) '''
gen test data
''' dnn_csvTEST = 'dnn_csvTEST{}.csv'.format(int(time.time()))
with open(dnn_csvTEST, 'w') as fa:
fa.write(csv_head)
with open(f_test, 'r') as fr:
for i in fr:
if 'aid' in i:
continue
try:
break
l = i.replace('\n', '').replace(' ', '').split(',')
print(l)
# aid, uid, label = l
aid, uid = l
label =
s = '{},{},{}\n'.format(label, uid_d[uid], ad_d[aid])
with open(dnn_csvTEST, 'a') as fa:
fa.write(s)
except Exception as e:
print(e) dd =
特征列 属性值 获取 vowpal wabbit 生成DNN 的训练测试数据的更多相关文章
- spring boot 配置属性值获取注解@Value和@ConfigurationProperties比较
功能比较 : @ConfigurationProperties @Value 映射赋值 批量注入配置文件中的属性 一个个指定 松散绑定(松散语法)① 支持 不支持 SpEL② 不支持 支持 ...
- SoapUI对于Json数据进行属性值获取与传递
SoapUI的Property Transfer功能可以很好地对接口请求返回的数据进行参数属性获取与传递,但对于Json数据,SoapUI会把数据格式先转换成XML格式,但实际情况却是,转换后的XML ...
- 【拼接属性查询方式】MySql某一列属性值为拼接时的查询方式
数据库中某一列的值为 使用IN查询是无法查询到数据的,只能查询到IN(2)或者IN(2,3) 查询结果: , style_id) 正确方式:使用FIND_IN_SET函数 , style_id) , ...
- 获取List集合对象中某一列属性值
例:获取disposeList集合中CorpusMarkPage对象中的responseId属性,生成新的List集合 List<String> responseIdList = disp ...
- XmlDocument根据节点的属性值获取节点
string targetParm = string.Format("STUDENTS/STUDENT[@NO='{0}']", targetValue);//生成目标获取节点的参 ...
- java类的静态属性值获取
获取某个类实例的静态属性: public class ErrorCode { private String code; private String message; private ErrorCod ...
- 在CSV文件中增加一列属性值
具体参见:系统管理\将文件夹复制到列表中的远程主机 修改前: column1, column2 1,b 2,c 3,5 修改后: column1, column2, column3 1,b, ...
- ContactsContract中涉及数据库中的一些列属性值【Written By KillerLegend】
ContactsContract.Data http://developer.android.com/reference/android/provider/ContactsContract.Dat ...
- ul标签中，li标签的移除、属性值获取
随机推荐
- 第3节 mapreduce高级：7、自定义outputformat实现输出到不同的文件夹下面
2.1 需求 现在有一些订单的评论数据,需求,将订单的好评与差评进行区分开来,将最终的数据分开到不同的文件夹下面去,数据内容参见资料文件夹,其中数据第九个字段表示好评,中评,差评.0:好评,1:中评, ...
- 从C#程序中调用非受管DLLs
从C#程序中调用非受管DLLs 文章概要: 众所周知,.NET已经渐渐成为一种技术时尚,那么C#很自然也成为一种编程时尚.如何利用浩如烟海的Win32 API以及以前所编写的 Win32 代码已经成为 ...
- 学习React从接受JSX开始
详情参考官方JSX规范 虽然JSX是扩展到ECMAScript的类XML语法,但是它本身并没有定义任何语义.也就是说它本身不在ECMAScript标准范围之内.它也不会被引擎或者浏览器直接执行.通常会 ...
- delphi clipboard
判断clipboard里的格式: if CliPBoard.HasFormat(CF_TEXT) then EdIT1.Text := CliPBoard.AsText ...
- 树莓派 -- oled 续（2） python
上文中的代码通过wiringPi的API调用devfs API来显示图片. 这里分析的Python代码也通过类似的方法来显示图片. 主要用到了两个Library. import spidev impo ...
- python 类的装饰器
我们知道,在不改变原有代码的基础上,我们可以使用装饰器为函数添加新的功能.同理,一切皆对象,我们也可以使用装饰器为类添加类属性.what? def deco(obj): obj.x = 1 obj.y ...
- LeetCode（59）SPiral Matrix II
题目 Given an integer n, generate a square matrix filled with elements from 1 to n2 in spiral order. F ...
- UVa 1599 理想路径（反向BFS 求最短路径 ）
题意: 给定一个有重边有自环的无向图,n个点(2 <= n <= 100000), m条边(1 <= m <= 200000), 每条边有一个权值, 求从第一个点到n的最少步数 ...
- Shader Wave
Shader Wave 一.原理 1. 采用 UV 坐标为原始数据,生成每一条波浪线. 2. 使用 Unity 的 Time.y 作为时间增量,动态变换波形. 二.操作步骤 1. 首先使用纹理坐标生成 ...
- hihoCoder #1162 : 骨牌覆盖问题·三
#1162 : 骨牌覆盖问题·三 Time Limit:10000ms Case Time Limit:1000ms Memory Limit:256MB 描述 前两周里,我们讲解了2xN,3xN骨牌 ...