/
external_kernel.py
177 lines (133 loc) · 6.17 KB
/
external_kernel.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 gc
import time
import numpy as np
import pandas as pd
from scipy.sparse import csr_matrix, hstack
from sklearn.linear_model import Ridge
from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer
from sklearn.preprocessing import LabelBinarizer
from sklearn.model_selection import train_test_split, cross_val_score
import lightgbm as lgb
NUM_BRANDS = 4000
NUM_CATEGORIES = 1000
NAME_MIN_DF = 10
MAX_FEATURES_ITEM_DESCRIPTION = 55000
def handle_missing_inplace(dataset):
dataset['category_name'].fillna(value='missing', inplace=True)
dataset['brand_name'].fillna(value='missing', inplace=True)
dataset['item_description'].fillna(value='missing', inplace=True)
def cutting(dataset):
pop_brand = dataset['brand_name'].value_counts().loc[lambda x: x.index != 'missing'].index[:NUM_BRANDS]
dataset.loc[~dataset['brand_name'].isin(pop_brand), 'brand_name'] = 'missing'
pop_category = dataset['category_name'].value_counts().loc[lambda x: x.index != 'missing'].index[:NUM_BRANDS]
dataset.loc[~dataset['category_name'].isin(pop_category), 'category_name'] = 'missing'
def to_categorical(dataset):
dataset['category_name'] = dataset['category_name'].astype('category')
dataset['brand_name'] = dataset['brand_name'].astype('category')
dataset['item_condition_id'] = dataset['item_condition_id'].astype('category')
def rmsle(y, y0):
print(y[0:10])
print(y0[0:10])
assert len(y) == len(y0)
return np.sqrt(np.mean(np.power(np.log1p(y)-np.log1p(y0), 2)))
def main():
start_time = time.time()
train = pd.read_table('../train.tsv', engine='c')
test = pd.read_table('../test.tsv', engine='c')
train = train.sample(frac=1).reset_index(drop=True)
test = train.loc[100000:120000]
train = train.loc[0:100000]
print('[{}] Finished to load data'.format(time.time() - start_time))
print('Train shape: ', train.shape)
print('Test shape: ', test.shape)
test_labels = np.log1p(test['price'])
nrow_train = train.shape[0]
# y = train["price"]
y = np.log1p(train["price"])
merge = pd.concat([train, test])
submission = test[['train_id']]
del train
del test
gc.collect()
handle_missing_inplace(merge)
print('[{}] Finished to handle missing'.format(time.time() - start_time))
cutting(merge)
print('[{}] Finished to cut'.format(time.time() - start_time))
to_categorical(merge)
print('[{}] Finished to convert categorical'.format(time.time() - start_time))
cv = CountVectorizer(min_df=NAME_MIN_DF)
X_name = cv.fit_transform(merge['name'])
print('[{}] Finished count vectorize `name`'.format(time.time() - start_time))
cv = CountVectorizer()
X_category = cv.fit_transform(merge['category_name'])
print('[{}] Finished count vectorize `category_name`'.format(time.time() - start_time))
tv = TfidfVectorizer(max_features=MAX_FEATURES_ITEM_DESCRIPTION,
ngram_range=(1, 3),
stop_words='english')
X_description = tv.fit_transform(merge['item_description'])
print('[{}] Finished TFIDF vectorize `item_description`'.format(time.time() - start_time))
lb = LabelBinarizer(sparse_output=True)
X_brand = lb.fit_transform(merge['brand_name'])
print('[{}] Finished label binarize `brand_name`'.format(time.time() - start_time))
X_dummies = csr_matrix(pd.get_dummies(merge[['item_condition_id', 'shipping']],
sparse=True).values)
print('[{}] Finished to get dummies on `item_condition_id` and `shipping`'.format(time.time() - start_time))
sparse_merge = hstack((X_dummies, X_description, X_brand, X_category, X_name)).tocsr()
print('[{}] Finished to create sparse merge'.format(time.time() - start_time))
X = sparse_merge[:nrow_train]
X_test = sparse_merge[nrow_train:]
print("TRAINING SHAPE")
print(X.shape)
print("Test SHAPE")
print(X_test.shape)
model = Ridge(solver="sag", fit_intercept=True, random_state=205, alpha=3)
model.fit(X, y)
print('[{}] Finished to train ridge sag'.format(time.time() - start_time))
predsR = model.predict(X=X_test)
print('[{}] Finished to predict ridge sag'.format(time.time() - start_time))
model = Ridge(solver="lsqr", fit_intercept=True, random_state=145, alpha = 3)
model.fit(X, y)
print('[{}] Finished to train ridge lsqrt'.format(time.time() - start_time))
predsR2 = model.predict(X=X_test)
print('[{}] Finished to predict ridge lsqrt'.format(time.time() - start_time))
train_X, valid_X, train_y, valid_y = train_test_split(X, y, test_size = 0.1, random_state = 144)
d_train = lgb.Dataset(train_X, label=train_y)
d_valid = lgb.Dataset(valid_X, label=valid_y)
watchlist = [d_train, d_valid]
params = {
'learning_rate': 0.76,
'application': 'regression',
'max_depth': 3,
'num_leaves': 99,
'verbosity': -1,
'metric': 'RMSE',
'nthread': 4
}
params2 = {
'learning_rate': 0.85,
'application': 'regression',
'max_depth': 3,
'num_leaves': 110,
'verbosity': -1,
'metric': 'RMSE',
'nthread': 4
}
model = lgb.train(params, train_set=d_train, num_boost_round=7500, valid_sets=watchlist, \
early_stopping_rounds=500, verbose_eval=500)
predsL = model.predict(X_test)
print('[{}] Finished to predict lgb 1'.format(time.time() - start_time))
train_X2, valid_X2, train_y2, valid_y2 = train_test_split(X, y, test_size = 0.1, random_state = 101)
d_train2 = lgb.Dataset(train_X2, label=train_y2)
d_valid2 = lgb.Dataset(valid_X2, label=valid_y2)
watchlist2 = [d_train2, d_valid2]
model = lgb.train(params2, train_set=d_train2, num_boost_round=3000, valid_sets=watchlist2, \
early_stopping_rounds=50, verbose_eval=500)
predsL2 = model.predict(X_test)
print('[{}] Finished to predict lgb 2'.format(time.time() - start_time))
preds = predsR2*0.15 + predsR*0.15 + predsL*0.5 + predsL2*0.2
submission['price'] = np.expm1(preds)
submission.to_csv("submission_lgbm_ridge_11.csv", index=False)
print("ERROR")
print(rmsle(preds, test_labels))
if __name__ == '__main__':
main()