-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
177 lines (150 loc) · 6.74 KB
/
utils.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
import pandas as pd
import numpy as np
from scipy.io import arff
import re
import sys
import os
from io import StringIO
def read_click(*args, **kwargs):
with open("./data/click/track2/subsampling_idx.txt") as fin:
ids = map(int, fin.readlines())
unique_ids = set(ids)
data_strings = {}
with open('./data/click/track2/training.txt') as fin:
for i, string in enumerate(fin):
if i in unique_ids:
data_strings[i] = string
data_rows = [v for _,v in data_strings.items()]
data = pd.read_table(StringIO("".join(data_rows)), header=None, delim_whitespace=True).apply(np.float64)
colnames = ['click',
'impression',
'url_hash',
'ad_id',
'advertiser_id',
'depth',
'position',
'query_id',
'keyword_id',
'title_id',
'description_id',
'user_id']
data.columns = colnames
data["Target"] = data["click"].apply(lambda x: 1 if x == 0 else -1)
data.drop(["click"], axis=1, inplace=True)
categorical_features = {1, 2, 3, 6, 7, 8, 9, 10}
def clean_string(s):
return "v_" + re.sub('[^A-Za-z0-9]+', "_", str(s))
for i in categorical_features:
data[data.columns[i]] = data[data.columns[i]].apply(clean_string)
data[data.columns[i]] = data[data.columns[i]].astype('category')
return data, []
def read_kick():
data = pd.read_csv("./data/kick/training.csv")
target = data["IsBadBuy"].apply(lambda x: 1.0 if x == 0 else -1.0)
data["PurchYear"] = pd.DatetimeIndex(data['PurchDate']).year
data["PurchMonth"] = pd.DatetimeIndex(data['PurchDate']).month
data["PurchDay"] = pd.DatetimeIndex(data['PurchDate']).day
data["PurchWeekday"] = pd.DatetimeIndex(data['PurchDate']).weekday
categorical_features = set([0, 1, 3, 4, 5, 6, 7, 8, 9, 10, 12, 13, 14, 23, 24, 25, 26, 27, 29, 31, 32, 33, 34])
data.drop(["RefId", "IsBadBuy", "PurchDate"], axis=1, inplace=True)
def clean_string(s):
return re.sub('[^A-Za-z0-9]+', "_", str(s))
for i in categorical_features:
data[data.columns[i]] = data[data.columns[i]].apply(clean_string)
columns_to_impute = []
for i, column in enumerate(data.columns):
if i not in categorical_features and pd.isnull(data[column]).any():
columns_to_impute.append(column)
for column_name in columns_to_impute:
data[column_name + "_imputed"] = pd.isnull(data[column_name]).astype(float)
data[column_name].fillna(0, inplace=True)
for i, column in enumerate(data.columns):
if i not in categorical_features:
data[column] = data[column].astype(float)
else:
data[column] = data[column].astype('category')
data["Target"] = target
return data, []
def read_adult(data_path='./data/adult/adult.data', names_path='./data/adult/adult.names'):
with open(names_path) as f:
lines = list(filter(lambda l: not l.startswith('|') and l.strip(), f.readlines()))
categorial = []
feature_name = []
for line in lines:
split = line.split(':')
if len(split) == 2:
categorial.append(not split[1].startswith('continuous'))
feature_name.append(split[0])
feature_name.append('Target')
categorial.append(True)
df = pd.read_csv(data_path, names=feature_name)
for n, c in zip(feature_name, categorial):
if c:
df[n] = df[n].astype('category').cat.codes
return df, [n for n,c in zip(feature_name, categorial) if c and n != 'Target']
def read_internet(data_path='./data/internet/kdd_internet_usage.arff'):
data, meta = arff.loadarff(data_path)
df = pd.DataFrame(data)
df['Target'] = df['Who_Pays_for_Access_Work'].apply(lambda x:int(x.decode()))
df.drop(["Who_Pays_for_Access_Work", "Willingness_to_Pay_Fees", "Years_on_Internet", "who"], axis=1, inplace=True)
for c in df.columns:
if c == 'Target':
continue
df[c] = df[c].apply(lambda x: x.decode()).astype('category')
return df, [c for c in df.columns if c !='Target']
def _prepare_kdd(data_path):
def to_float_str(element):
try:
return str(float(element))
except ValueError:
return element
data = pd.read_csv(data_path, sep='\t')
categorical_features = { 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228 }
for i in categorical_features:
data[data.columns[i]].fillna("?", inplace=True)
data[data.columns[i]] = data[data.columns[i]].astype('category')
columns_to_impute = []
for i, column in enumerate(data.columns):
if i not in categorical_features and pd.isnull(data[column]).any():
columns_to_impute.append(column)
for column_name in columns_to_impute:
data[column_name].fillna(0, inplace=True)
for i, column in enumerate(data.columns):
if i not in categorical_features:
data[column] = data[column].astype(float)
return data
def read_appet(data_path='./data/appet/orange_small_train.data/orange_small_train.data',
label_path='./data/appet/orange_small_train.data/orange_small_train_appetency.labels'):
data = _prepare_kdd(data_path)
with open(label_path) as f:
data["Target"] = list(map(lambda x: (int(x) + 1)/2, f.readlines()))
return data, []
def read_kddchurn(data_path='./data/appet/orange_small_train.data/orange_small_train.data',
label_path='./data/appet/orange_small_train.data/orange_large_train_churn.labels'):
data = _prepare_kdd(data_path)
with open(label_path) as f:
data["Target"] = list(map(lambda x: (int(x) + 1)/2, f.readlines()))
return data, []
def read_upsel(data_path='./data/appet/orange_small_train.data/orange_small_train.data',
label_path='./data/appet/orange_small_train.data/orange_small_train_upselling.labels'):
data = _prepare_kdd(data_path)
with open(label_path) as f:
data["Target"] = list(map(lambda x: (int(x) + 1)/2, f.readlines()))
return data, []
def read_amazon(data_path='./data/amazon/train.csv'):
df = pd.read_csv(data_path)
print([(c, df[c].nunique()) for c in df.columns])
df['Target'] = df['ACTION']
df.drop(["ACTION", "RESOURCE"], axis=1, inplace=True)
return df, []
def read_dataset_by_name(dataset_name, *args, **kwargs):
return {
'adult': read_adult,
'internet': read_internet,
'amazon': read_amazon,
# 'appet' : read_appet,
# 'kddchurn': read_kddchurn,
# 'upsel': read_upsel
'click': read_click,
'kick': read_kick
}[dataset_name](*args, **kwargs)