df = transform(df)
os.makedirs(output_dir, exist_ok=True)
params = {
'objective': 'binary',
'num_leaves': 32,
'min_data_in_leaf': 15, # 42,
'max_depth': -1,
'learning_rate': 0.1,
'boosting': 'gbdt',
'bagging_fraction': 0.7, # 0.5,
'feature_fraction': 0.5,
'bagging_seed': 0,
'reg_alpha': 0.1, # 1.728910519108444,
'reg_lambda': 1,
'random_state': 0,
'metric': 'auc',
'verbosity': -1,
"n_estimators": 10000,
"early_stopping_rounds": 100
}
df = df.drop(["user_answer"], axis=1)
train_lgbm_cv(df,
params=params,
output_dir=output_dir,
model_id=model_id,
exp_name="exp003")
break
'verbosity': -1,
"n_estimators": 10000,
"early_stopping_rounds": 50
}
df.tail(1000).to_csv("exp028.csv", index=False)
df = df.drop(["user_answer", "tags", "type_of"], axis=1)
df = df[df["answered_correctly"].notnull()]
print(df.columns)
print(df.shape)
print(model_id)
train_lgbm_cv(df,
params=params,
output_dir=output_dir,
model_id=model_id,
exp_name=model_id,
is_debug=is_debug,
drop_user_id=True)
params = {
'n_estimators': 12000,
'learning_rate': 0.1,
'eval_metric': 'AUC',
'loss_function': 'Logloss',
'random_seed': 0,
'metric_period': 50,
'od_wait': 400,
'task_type': 'GPU',
'max_depth': 8,
"verbose": 100
# df.to_pickle("../input/riiid-test-answer-prediction/train_merged.pickle")
# print(len(df))
# df = pd.read_pickle("../input/riiid-test-answer-prediction/train_merged.pickle")
df = transform(df)
os.makedirs(output_dir, exist_ok=True)
params = {
'objective': 'binary',
'num_leaves': 32,
'min_data_in_leaf': 15, # 42,
'max_depth': -1,
'learning_rate': 0.1,
'boosting': 'gbdt',
'bagging_fraction': 0.7, # 0.5,
'feature_fraction': 0.5,
'bagging_seed': 0,
'reg_alpha': 0.1, # 1.728910519108444,
'reg_lambda': 1,
'random_state': 0,
'metric': 'auc',
'verbosity': -1,
"n_estimators": 10000,
"early_stopping_rounds": 100
}
df = df.drop(["user_answer", "row_id"], axis=1)
train_lgbm_cv(df, params=params, output_dir=output_dir, model_id=model_id)
break
'objective': 'binary',
'num_leaves': 32,
'min_data_in_leaf': 15, # 42,
'max_depth': -1,
'learning_rate': 0.1,
'boosting': 'gbdt',
'bagging_fraction': 0.7, # 0.5,
'feature_fraction': 0.5,
'bagging_seed': 0,
'reg_alpha': 0.1, # 1.728910519108444,
'reg_lambda': 1,
'random_state': 0,
'metric': 'auc',
'verbosity': -1,
"n_estimators": 10000,
"early_stopping_rounds": 100
}
df = df.drop(["user_answer"], axis=1)
print(df.columns)
model_id = os.path.basename(fname).replace(".pickle", "")
print(model_id)
train_lgbm_cv(df,
params=params,
output_dir=output_dir,
model_id=model_id,
exp_name=f"exp008_{model_id}",
drop_user_id=True)
break
print(model_id)
for _ in range(10000):
params = {
'objective': 'binary',
'num_leaves': random.choice([8, 16, 32, 64, 128]),
'max_depth': -1,
'learning_rate': 0.3,
'boosting': random.choice(['gbdt', 'gbdt', 'gbdt', 'goss']),
'bagging_fraction': random.choice([0.1, 0.5, 0.7, 0.9]),
'feature_fraction': random.choice([0.1, 0.3, 0.5, 0.7, 0.9]),
'bagging_seed': 0,
'reg_alpha': random.choice([0, 0.1, 1, 5]),
'reg_lambda': random.choice([0, 0.1, 1, 5]),
'random_state': 0,
'metric': 'auc',
'verbosity': -1,
"n_estimators": 10000,
"early_stopping_rounds": 100
}
model_id = os.path.basename(fname).replace(".pickle", "")
train_lgbm_cv(df,
params=params,
output_dir=output_dir,
model_id=model_id,
exp_name=model_id,
drop_user_id=True,
experiment_id=2)
del params
break
params = {
'objective': 'binary',
'num_leaves': 32,
'min_data_in_leaf': 15, # 42,
'max_depth': -1,
'learning_rate': 0.3,
'boosting': 'gbdt',
'bagging_fraction': 0.7, # 0.5,
'feature_fraction': 0.9,
'bagging_seed': 0,
'reg_alpha': 5, # 1.728910519108444,
'reg_lambda': 5,
'random_state': 0,
'metric': 'auc',
'verbosity': -1,
"n_estimators": 10000,
"early_stopping_rounds": 100
}
df = df.drop(["user_answer", "tags", "type_of"], axis=1)
df = df[df["answered_correctly"].notnull()]
print(df.columns)
print(df.shape)
train_lgbm_cv(df,
params=params,
output_dir=output_dir,
model_id=0,
exp_name="exp11",
drop_user_id=True)
"n_estimators": 10000,
"early_stopping_rounds": 50
}
df.tail(1000).to_csv("exp028.csv", index=False)
df = df[df["answered_correctly"].notnull()]
print("lgbm")
print(df.columns)
print(df.shape)
model_id = os.path.basename(fname).replace(".pickle", "")
print(model_id)
train_lgbm_cv(df.drop(["user_answer", "tags", "type_of"], axis=1),
params=params,
output_dir=output_dir,
model_id=model_id,
exp_name=model_id,
is_debug=is_debug,
drop_user_id=True)
useful_cols = [
"target_enc_content_id",
"target_enc_['content_id', 'prior_question_had_explanation']",
"user_rate_mean_content_id", "user_rate_sum_content_id",
"shiftdiff_timestamp_by_user_id",
"previous_answer_['user_id', 'content_id']",
"target_enc_['content_id', 'user_count_bin']",
"target_enc_['user_id', 'part']",
"diff_user_level_target_enc_content_id", "user_id",
"answered_correctly"
]
