from pipeline import Pipe
from model import xgboost_model, lightgbm_model
from sklearn.cross_validation import train_test_split
import xgboost as xgb
import lightgbm as lgb
if __name__ == '__main__':
train_path = 'data/train_test.csv'
val_path = 'data/train_ensemble.csv'
test_path = 'data/test_clean.csv'
firsttime = True
if firsttime:
training_pipe = Pipe(train_path, val_path)
train = training_pipe.make('pickle/train')
train.to_csv('train_feature.csv', index=False)
testing_pipe = Pipe(val_path, test_path)
test = testing_pipe.make('pickle/test')
test.to_csv('test_features.csv', index=False)
else:
train = pd.read_csv('train_feature.csv')
test = pd.read_csv('test_features.csv')
featlist = train.columns.tolist()
featlist.remove('is_listened')
X = train[featlist].as_matrix()
y = train['is_listened'].as_matrix()
test_X = test[featlist].as_matrix()
if __name__ == '__main__':
# Input files
train_path = 'data/archive/train_clean.csv'
test_path = 'data/archive/test_clean.csv'
# Intermediate files
TRAIN_PATH_INTERMEDIATE = 'data/archive/train_intermediate.csv'
TEST_PATH_INTERMEDIATE = 'data/archive/test_intermediate.csv'
# Set boolean to use pre-made features or build on the fly
firsttime = True
if firsttime:
training_pipe = Pipe(train_path, train_path)
train = training_pipe.make('train_intermediate')
train.to_csv(TRAIN_PATH_INTERMEDIATE, index=False)
testing_pipe = Pipe(train_path, test_path)
test = testing_pipe.make('test_intermediate')
test.to_csv(TEST_PATH_INTERMEDIATE, index=False)
else:
train = pd.read_csv('train_feature.csv')
test = pd.read_csv('test_features.csv')
# Preparing train and test dataset for ensemble layer
featlist = train.columns.tolist()
featlist.remove('is_listened')
X = train[featlist].as_matrix()
y = train['is_listened'].as_matrix()