-
Notifications
You must be signed in to change notification settings - Fork 19
/
Copy pathgenerate_loader.py
executable file
·183 lines (152 loc) · 7.7 KB
/
generate_loader.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
import numpy as np
import pandas as pd
import pickle
import copy
import os
from xgboost import XGBClassifier
from utils import find_best_threshold,process_leaf_idx,stratify_sample
from sklearn.metrics import f1_score,roc_auc_score
import torch
import torch.utils.data as Data
import warnings
warnings.filterwarnings("ignore")
# load preprocessed data
with open("./processed_data.pickle","rb") as f :
processed_data = pickle.load(f)
print(processed_data.keys())
print("Finish loading data...")
# train/test data
train = processed_data["raw"]["train"]
valid = processed_data["raw"]["valid"]
test = processed_data["raw"]["test"]
# Revenue data for regression target
revenue_train, revenue_valid,revenue_test = processed_data["revenue"]["train"],\
processed_data["revenue"]["valid"],\
processed_data["revenue"]["test"]
# normalize revenue by f(x) = log(x+1)/max(xi)
norm_revenue_train, norm_revenue_test = np.log(revenue_train+1), np.log(revenue_test+1)
global_max = max(norm_revenue_train)
norm_revenue_train = norm_revenue_train/global_max
# Xgboost data
xgb_trainx = processed_data["xgboost_data"]["train_x"]
xgb_trainy = processed_data["xgboost_data"]["train_y"]
xgb_validx = processed_data["xgboost_data"]["valid_x"]
xgb_validy = processed_data["xgboost_data"]["valid_y"]
xgb_testx = processed_data["xgboost_data"]["test_x"]
xgb_testy = processed_data["xgboost_data"]["test_y"]
# build xgboost model
print("Training xgboost model...")
columns = ['fob.value', 'cif.value', 'total.taxes', 'gross.weight', 'quantity', 'Unitprice', 'WUnitprice', 'TaxRatio', 'FOBCIFRatio', 'TaxUnitquantity', 'tariff.code', 'HS6', 'HS4', 'HS2', 'SGD.DayofYear', 'SGD.WeekofYear', 'SGD.MonthofYear'] + [col for col in train.columns if 'RiskH' in col]
xgb_trainx = pd.DataFrame(xgb_trainx,columns=columns)
xgb_validx = pd.DataFrame(xgb_validx,columns=columns)
xgb_testx = pd.DataFrame(xgb_testx,columns=columns)
xgb_clf = XGBClassifier(n_estimators=100, max_depth=4,n_jobs=-1)
xgb_clf.fit(xgb_trainx,xgb_trainy)
# evaluate xgboost model
print("------Evaluating xgboost model------")
test_pred = xgb_clf.predict_proba(xgb_testx)[:,1]
xgb_auc = roc_auc_score(xgb_testy, test_pred)
xgb_threshold,_ = find_best_threshold(xgb_clf, xgb_validx, xgb_validy)
xgb_f1 = find_best_threshold(xgb_clf, xgb_testx, xgb_testy,best_thresh=xgb_threshold)
print("AUC = %.4f, F1-score = %.4f" % (xgb_auc, xgb_f1))
# Precision and Recall
y_prob = test_pred
for i in [99,98,95,90]:
threshold = np.percentile(y_prob, i)
print(f'Checking top {100-i}% suspicious transactions: {len(y_prob[y_prob > threshold])}')
precision = np.mean(xgb_testy[y_prob > threshold])
recall = sum(xgb_testy[y_prob > threshold])/sum(xgb_testy)
revenue_recall = sum(revenue_test[y_prob > threshold]) /sum(revenue_test)
print(f'Precision: {round(precision, 4)}, Recall: {round(recall, 4)}, Seized Revenue (Recall): {round(revenue_recall, 4)}')
xgb_clf.get_booster().dump_model('xgb_model.txt', with_stats=False)
# Xgboost+LR model
from sklearn.linear_model import LogisticRegression
from sklearn.preprocessing import OneHotEncoder
# get leaf index from xgboost model
X_train_leaves = xgb_clf.apply(xgb_trainx)
X_valid_leaves = xgb_clf.apply(xgb_validx)
X_test_leaves = xgb_clf.apply(xgb_testx)
train_rows = X_train_leaves.shape[0]
# one-hot encoding for leaf index
xgbenc = OneHotEncoder(categories="auto")
lr_trainx = xgbenc.fit_transform(X_train_leaves)
lr_validx = xgbenc.transform(X_valid_leaves)
lr_testx = xgbenc.transform(X_test_leaves)
# model
print("Training Logistic regression model...")
lr = LogisticRegression(n_jobs=-1)
lr.fit(lr_trainx, xgb_trainy)
test_pred = lr.predict_proba(lr_testx)[:,1]
print("------Evaluating xgboost+LR model------")
xgb_auc = roc_auc_score(xgb_testy, test_pred)
xgb_threshold,_ = find_best_threshold(lr, lr_validx, xgb_validy) # threshold was select from validation set
xgb_f1 = find_best_threshold(lr, lr_testx, xgb_testy,best_thresh=xgb_threshold) # then applied on test set
print("AUC = %.4f, F1-score = %.4f" % (xgb_auc, xgb_f1))
# Precision and Recall
y_prob = test_pred
for i in [99,98,95,90]:
threshold = np.percentile(y_prob, i)
print(f'Checking top {100-i}% suspicious transactions: {len(y_prob[y_prob > threshold])}')
precision = np.mean(xgb_testy[y_prob > threshold])
recall = sum(xgb_testy[y_prob > threshold])/sum(xgb_testy)
revenue_recall = sum(revenue_test[y_prob > threshold]) /sum(revenue_test)
print(f'Precision: {round(precision, 4)}, Recall: {round(recall, 4)}, Seized Revenue (Recall): {round(revenue_recall, 4)}')
# user & item information
train_raw_importers = train['importer.id'].values
train_raw_items = train['tariff.code'].values
valid_raw_importers = valid['importer.id'].values
valid_raw_items = valid['tariff.code'].values
test_raw_importers = test['importer.id']
test_raw_items = test['tariff.code']
# we need padding for unseen user or item
importer_set = set(train_raw_importers)
item_set = set(train_raw_items)
# Remember to +1 for zero padding
importer_mapping = {v:i+1 for i,v in enumerate(importer_set)}
hs6_mapping = {v:i+1 for i,v in enumerate(item_set)}
importer_size = len(importer_mapping) + 1
item_size = len(hs6_mapping) + 1
train_importers = [importer_mapping[x] for x in train_raw_importers]
train_items = [hs6_mapping[x] for x in train_raw_items]
# for test data, we use padding_idx=0 for unseen data
valid_importers = [importer_mapping.get(x,0) for x in valid_raw_importers]
valid_items = [hs6_mapping.get(x,0) for x in valid_raw_items]
test_importers = [importer_mapping.get(x,0) for x in test_raw_importers] # use dic.get(key,deafault) to handle unseen
test_items = [hs6_mapping.get(x,0) for x in test_raw_items]
# Preprocess
train_rows = train.shape[0]
valid_rows = valid.shape[0] + train_rows
X_leaves = np.concatenate((X_train_leaves, X_valid_leaves, X_test_leaves), axis=0) # make sure the dimensionality
transformed_leaves, leaf_num, new_leaf_index = process_leaf_idx(X_leaves)
train_leaves, valid_leaves, test_leaves = transformed_leaves[:train_rows],\
transformed_leaves[train_rows:valid_rows],\
transformed_leaves[valid_rows:]
# Convert to torch type
train_leaves = torch.tensor(train_leaves).long()
train_user = torch.tensor(train_importers).long()
train_item = torch.tensor(train_items).long()
valid_leaves = torch.tensor(valid_leaves).long()
valid_user = torch.tensor(valid_importers).long()
valid_item = torch.tensor(valid_items).long()
test_leaves = torch.tensor(test_leaves).long()
test_user = torch.tensor(test_importers).long()
test_item = torch.tensor(test_items).long()
# cls data
train_label_cls = torch.tensor(xgb_trainy).float()
valid_label_cls = torch.tensor(xgb_validy).float()
test_label_cls = torch.tensor(xgb_testy).float()
# revenue data
train_label_reg = torch.tensor(norm_revenue_train).float()
valid_label_reg = torch.tensor(revenue_valid).float()
test_label_reg = torch.tensor(revenue_test).float()
# create dataloader
train_dataset = Data.TensorDataset(train_leaves,train_user,train_item,train_label_cls,train_label_reg)
valid_dataset = Data.TensorDataset(valid_leaves,valid_user,valid_item,valid_label_cls,valid_label_reg)
test_dataset = Data.TensorDataset(test_leaves,test_user,test_item,test_label_cls,test_label_reg)
data4embedding = {"train_dataset":train_dataset,"valid_dataset":valid_dataset,"test_dataset":test_dataset,\
"leaf_num":leaf_num,"importer_num":importer_size,"item_size":item_size}
# save data
with open("torch_data.pickle", 'wb') as f:
pickle.dump(data4embedding, f, protocol=pickle.HIGHEST_PROTOCOL)
with open("leaf_index.pickle", "wb") as f:
pickle.dump(new_leaf_index, f, protocol=pickle.HIGHEST_PROTOCOL)