-
Notifications
You must be signed in to change notification settings - Fork 1
/
model_training.py
173 lines (129 loc) · 5.25 KB
/
model_training.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
from audioop import add
import pandas as pd
import pickle
import warnings
warnings.filterwarnings("ignore")
from sklearn.feature_extraction import DictVectorizer
from sklearn.linear_model import LinearRegression, Lasso, Ridge
from sklearn.metrics import mean_squared_error
import xgboost as xgb
from hyperopt import fmin, tpe, hp, STATUS_OK, Trials
from hyperopt.pyll import scope
import mlflow
mlflow.set_tracking_uri("sqlite:///mlflow.db")
mlflow.set_experiment("nyc-taxi-experiment")
def read_dataframe(filename):
df = pd.read_parquet(filename)
df.lpep_dropoff_datetime = pd.to_datetime(df.lpep_dropoff_datetime)
df.lpep_pickup_datetime = pd.to_datetime(df.lpep_pickup_datetime)
df['duration'] = df.lpep_dropoff_datetime - df.lpep_pickup_datetime
df.duration = df.duration.apply(lambda td: td.total_seconds() / 60)
df = df[(df.duration >= 1) & (df.duration <= 60)]
categorical = ['PULocationID', 'DOLocationID']
df[categorical] = df[categorical].astype(str)
return df
def add_features(train_path="./data/green_tripdata_2021-01.parquet",
val_path="./data/green_tripdata_2021-02.parquet"):
df_train = read_dataframe(train_path)
df_val = read_dataframe(val_path)
print(len(df_train))
print(len(df_val))
df_train['PU_DO'] = df_train['PULocationID'] + '_' + df_train['DOLocationID']
df_val['PU_DO'] = df_val['PULocationID'] + '_' + df_val['DOLocationID']
categorical = ['PU_DO'] #'PULocationID', 'DOLocationID']
numerical = ['trip_distance']
dv = DictVectorizer()
train_dicts = df_train[categorical + numerical].to_dict(orient='records')
X_train = dv.fit_transform(train_dicts)
val_dicts = df_val[categorical + numerical].to_dict(orient='records')
X_val = dv.transform(val_dicts)
target = 'duration'
y_train = df_train[target].values
y_val = df_val[target].values
return X_train, X_val, y_train, y_val, dv
# # Modelling
# lr = LinearRegression()
# lr.fit(X_train, y_train)
# y_pred = lr.predict(X_val)
# mean_squared_error(y_val, y_pred, squared=False)
# with open('models/lin_reg.bin', 'wb') as f_out:
# pickle.dump((dv, lr), f_out)
# with mlflow.start_run():
# mlflow.set_tag("developer", "cristian")
# mlflow.log_param("train-data-path", "./data/green_tripdata_2021-01.csv")
# mlflow.log_param("valid-data-path", "./data/green_tripdata_2021-02.csv")
# alpha = 0.1
# mlflow.log_param("alpha", alpha)
# lr = Lasso(alpha)
# lr.fit(X_train, y_train)
# y_pred = lr.predict(X_val)
# rmse = mean_squared_error(y_val, y_pred, squared=False)
# mlflow.log_metric("rmse", rmse)
# mlflow.log_artifact(local_path="models/lin_reg.bin", artifact_path="models_pickle")
def train_model_search(train, valid, y_val):
def objective(params):
with mlflow.start_run():
mlflow.set_tag("model", "xgboost")
mlflow.log_params(params)
booster = xgb.train(
params=params,
dtrain=train,
num_boost_round=1000,
evals=[(valid, 'validation')],
early_stopping_rounds=50
)
y_pred = booster.predict(valid)
rmse = mean_squared_error(y_val, y_pred, squared=False)
mlflow.log_metric("rmse", rmse)
return {'loss': rmse, 'status': STATUS_OK}
search_space = {
'max_depth': scope.int(hp.quniform('max_depth', 4, 100, 1)),
'learning_rate': hp.loguniform('learning_rate', -3, 0),
'reg_alpha': hp.loguniform('reg_alpha', -5, -1),
'reg_lambda': hp.loguniform('reg_lambda', -6, -1),
'min_child_weight': hp.loguniform('min_child_weight', -1, 3),
'objective': 'reg:linear',
'seed': 42
}
best_result = fmin(
fn=objective,
space=search_space,
algo=tpe.suggest,
max_evals=1,
trials=Trials()
)
return
def train_best_model(train, valid, y_val, dv):
with mlflow.start_run():
train = xgb.DMatrix(X_train, label=y_train)
valid = xgb.DMatrix(X_val, label=y_val)
best_params = {
'learning_rate': 0.09585355369315604,
'max_depth': 30,
'min_child_weight': 1.060597050922164,
'objective': 'reg:linear',
'reg_alpha': 0.018060244040060163,
'reg_lambda': 0.011658731377413597,
'seed': 42
}
mlflow.log_params(best_params)
booster = xgb.train(
params=best_params,
dtrain=train,
num_boost_round=1000,
evals=[(valid, 'validation')],
early_stopping_rounds=50
)
y_pred = booster.predict(valid)
rmse = mean_squared_error(y_val, y_pred, squared=False)
mlflow.log_metric("rmse", rmse)
with open("models/preprocessor.b", "wb") as f_out:
pickle.dump(dv, f_out)
mlflow.log_artifact("models/preprocessor.b", artifact_path="preprocessor")
mlflow.xgboost.log_model(booster, artifact_path="models_mlflow")
if __name__ == "__main__":
X_train, X_val, y_train, y_val, dv = add_features()
train = xgb.DMatrix(X_train, label=y_train)
valid = xgb.DMatrix(X_val, label=y_val)
train_model_search(train, valid, y_val)
train_best_model(train, valid, y_val, dv)