def read_data(self):
data_set = Dataset(self.train_file_name,
self.test_file_name,
base_dir=self.data_dir)
self.train_X, self.train_Y, self.test, self.train_features, self.cate_features = data_set.preprocess(
reload=True)
def read_data(self):
self.set_train_test_bool = True
data_set = Dataset(self.train_file_name,
self.test_file_name,
base_dir=self.data_dir)
self.train_X, self.train_Y, self.test, self.train_features, self.cate_features = data_set.preprocess(
reload=True)
# default stratified values
self.stratified_values = self.train_Y.values
def set_train_outlier(self):
dataset = Dataset(train_path=self.train_file, test_path=self.test_file)
self.train_X, self.train_Y, self.test, self.features, self.cate_features = dataset.preprocess(
reload=True)
self.train_X['target'] = self.train_Y
if 'outliers' not in self.train_X.columns:
dataset.set_outlier_col(self.train_X)
train_df = self.train_X[self.train_X['outliers'] == 0]
target = train_df['target']
del train_df['target']
return train_df, target
def load_data(self):
dataset = Dataset('train_agg_id1.csv', 'test_agg_id1.csv')
self.train_X, self.target, self.test, self.features, self.cate_features = dataset.preprocess(
reload=True)
param_dict[key] = params[index]
# set some non-numeric parameters too
for key in self.non_numeric_param:
param_dict[key] = self.non_numeric_param[key]
print(param_dict)
cv_error, _ = self._train(param_dict)
return cv_error, param_dict
def predict(self):
pass
if __name__ == "__main__":
dataset = Dataset(train_path='df_train_agg1.csv',
test_path='df_test_agg1.csv')
dataset.format_transformer(
train_file_name='alltrainffm_agg1.txt',
test_file_name='alltestffm_agg1.txt',
fields=[
'feature_1', 'feature_2', 'feature_3', 'elapsed_time',
'hist_month_lag_max', 'hist_category_1_sum', 'hist_weekend_sum',
'hist_category_3_mean_mean', 'hist_category_1_sum',
'hist_category_1_mean', 'hist_authorized_flag_sum',
'hist_authorized_flag_mean', 'hist_purchase_date_max'
])
model = FFMModel(debug=True,
train_name='alltrainffm_agg1.txt',
test_name='alltestffm_agg1.txt')
model.train()