rov-offline/params_search.py

import pandas as pd
import lightgbm as lgb
from sklearn.model_selection import train_test_split
from sklearn.model_selection import GridSearchCV
from sklearn.metrics import mean_absolute_percentage_error, r2_score
from rov_train import process_data
from config import set_config

config_, env = set_config()

train_filename = config_.TRAIN_DATA_FILENAME
X, Y, videos, fea = process_data(filename=train_filename)
x_train, x_test, y_train, y_test = train_test_split(X, Y, test_size=0.33)

# params = {
#           'boosting_type': 'gbdt',
#           'objective': 'regression',
#           'metric': 'mape',
#           'nthread': 4,
#           'learning_rate': 0.1,
#           'num_leaves': 30,
#           'max_depth': 5,
#           'subsample': 0.8,
#           'colsample_bytree': 0.8,
#     }
# data_train = lgb.Dataset(x_train, y_train)
# cv_results = lgb.cv(
#     params,
#     data_train,
#     num_boost_round=1000,
#     nfold=5,
#     stratified=False,
#     shuffle=True,
#     metrics='mape',
#     early_stopping_rounds=50,
#     seed=0
# )
# print('cv_results: ', cv_results)
# print('best n_estimators: ', len(cv_results['mape-mean']))
# print('best cv score: ', pd.Series(cv_results['mape-mean']).min())

# best n_estimators:  1000
# best cv score:  0.1578205729958749


# params_test1 = {'max_depth': range(3, 8, 1), 'num_leaves': range(5, 100, 10)}
# gsearch1 = GridSearchCV(
#     estimator=lgb.LGBMRegressor(boosting_type='gbdt', objective='regression', metrics='mape', learning_rate=0.1,
#                                 n_estimators=1000, max_depth=6, bagging_fraction=0.8, feature_fraction=0.8),
#     param_grid=params_test1, scoring='r2', cv=5, n_jobs=-1)
# gsearch1.fit(x_train, y_train)
# print(gsearch1.cv_results_, gsearch1.best_params_, gsearch1.best_score_)

# {'max_depth': 3, 'num_leaves': 5} 0.9263654009553877


# params_test2 = {'max_bin': range(5, 256, 10), 'min_data_in_leaf': range(1, 102, 10)}
#
# gsearch2 = GridSearchCV(
#     estimator=lgb.LGBMRegressor(boosting_type='gbdt', objective='regression', metrics='mape', learning_rate=0.1,
#                                 n_estimators=1000, max_depth=3, num_leaves=5, bagging_fraction=0.8,
#                                 feature_fraction=0.8),
#     param_grid=params_test2, scoring='r2', cv=5, n_jobs=-1)
# gsearch2.fit(x_train, y_train)
# print(gsearch2.cv_results_, gsearch2.best_params_, gsearch2.best_score_)

# {'max_bin': 125, 'min_data_in_leaf': 21} 0.7918455469932647


# params_test3 = {'feature_fraction': [0.6, 0.7, 0.8, 0.9, 1.0],
#                 'bagging_fraction': [0.6, 0.7, 0.8, 0.9, 1.0],
#                 'bagging_freq': range(0, 81, 10)
#                 }
#
# gsearch3 = GridSearchCV(
#     estimator=lgb.LGBMRegressor(boosting_type='gbdt', objective='regression', metrics='mape', learning_rate=0.1,
#                                 n_estimators=1000, max_depth=3, num_leaves=5, max_bin=125, min_data_in_leaf=21),
#     param_grid=params_test3, scoring='r2', cv=5, n_jobs=-1)
# gsearch3.fit(x_train, y_train)
# print(gsearch3.cv_results_, gsearch3.best_params_, gsearch3.best_score_)

# {'bagging_fraction': 0.6, 'bagging_freq': 0, 'feature_fraction': 0.9} 0.9003939890191465


# params_test4 = {'lambda_l1': [1e-5, 1e-3, 1e-1, 0.0, 0.1, 0.3, 0.5, 0.7, 0.9, 1.0],
#                 'lambda_l2': [1e-5, 1e-3, 1e-1, 0.0, 0.1, 0.3, 0.5, 0.7, 0.9, 1.0]
#                 }
#
# gsearch4 = GridSearchCV(
#     estimator=lgb.LGBMRegressor(boosting_type='gbdt', objective='regression', metrics='mape', learning_rate=0.1,
#                                 n_estimators=1000, max_depth=3, num_leaves=5, max_bin=125, min_data_in_leaf=21,
#                                 bagging_fraction=0.6, bagging_freq=0, feature_fraction=0.9),
#     param_grid=params_test4, scoring='r2', cv=5, n_jobs=-1)
# gsearch4.fit(x_train, y_train)
# print(gsearch4.cv_results_, gsearch4.best_params_, gsearch4.best_score_)

# {'lambda_l1': 0.7, 'lambda_l2': 0.9} 0.7538021411131314


# params_test5 = {'min_split_gain': [0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0]}
# gsearch5 = GridSearchCV(
#     estimator=lgb.LGBMRegressor(boosting_type='gbdt', objective='regression', metrics='mape', learning_rate=0.1,
#                                 n_estimators=1000, max_depth=3, num_leaves=5, max_bin=125, min_data_in_leaf=21,
#                                 bagging_fraction=0.6, bagging_freq=0, feature_fraction=0.9,
#                                 lambda_l1=0.7, lambda_l2=0.9),
#     param_grid=params_test5, scoring='r2', cv=5, n_jobs=-1)
# gsearch5.fit(x_train, y_train)
# print(gsearch5.cv_results_, gsearch5.best_params_, gsearch5.best_score_)
#
# {'min_split_gain': 0.0} 0.8903744931054016


# model = lgb.LGBMRegressor(boosting_type='gbdt', objective='regression', metrics='mape', learning_rate=0.01,
#                           n_estimators=7000, max_depth=3, num_leaves=5, max_bin=125, min_data_in_leaf=21,
#                           bagging_fraction=0.6, bagging_freq=0, feature_fraction=0.9,
#                           lambda_l1=0.7, lambda_l2=0.9, min_split_gain=0)
# model.fit(x_train, y_train)
# y_pre = model.predict(x_test)
# print('mape: ', mean_absolute_percentage_error(y_test, y_pre))
# print('r2: ', r2_score(y_test, y_pre))

# mape:  0.7407816559793686
# r2:  0.9069208480190998


# model = lgb.LGBMRegressor()
# model.fit(x_train, y_train)
# y_pre = model.predict(x_test)
# print('mape: ', mean_absolute_percentage_error(y_test, y_pre))
# print('r2: ', r2_score(y_test, y_pre))

# mape:  0.211698833035155
# r2:  0.893979193478978


model = lgb.LGBMRegressor(boosting_type='gbdt', objective='regression', metrics='mape', learning_rate=0.08,
                          n_estimators=3000, max_depth=3, num_leaves=5, max_bin=125, min_data_in_leaf=21,
                          bagging_fraction=0.6, bagging_freq=0, feature_fraction=0.9,
                          lambda_l1=0.7, lambda_l2=0.9, min_split_gain=0)
model.fit(x_train, y_train)
y_pre = model.predict(x_test)
print('mape: ', mean_absolute_percentage_error(y_test, y_pre))
print('r2: ', r2_score(y_test, y_pre))