def ridge_submit():
logger = mylogger()
logger.info('RidgeRegression start')
logger.debug('make_train_data start')
#train = pd.read_csv('../result_tmp/scaled_train.csv')
train = pd.read_csv('../result_tmp/scaled_train_DateBlockNum.csv')
#train = train[train['date_block_num']==33] #直近1ヶ月
train = train.loc[(30<train['date_block_num'])&(train['date_block_num']<=33)] #直近3m
y = train['item_cnt_month']
X = train.drop(['item_cnt_month', 'date_block_num'], axis=1).values
#X = train.drop(['item_cnt_month'], axis=1).values
#X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=1)
logger.debug('make_train_data end')
logger.info('Fitting start')
ridge = Ridge()
ridge.fit(X, y)
logger.debug('Fitting end')
logger.info('Scoring start')
#logger.info('Accuracy on test set: {:.3f}'.format(.score(X_test, y_test)))
test_data = load_test_data()
test = test_data.drop(['ID'], axis=1).values
submission = load_submission()
submission['item_cnt_month'] = ridge.predict(test).astype(np.float16).clip(0., 20.)
submission.to_csv('../result_tmp/submit_180902_31-33_ridge.csv', encoding='utf-8-sig', index=False)
logger.info('submission:\n{}'.format(submission.head()))
logger.debug('RidgeRegression end')
logger.debug('====================')
def RandomForest():
logger.info('RandomForestRegressor start')
logger.debug('make_train_data start')
train = pd.read_csv('./result_tmp/scaled_train.csv')
y = train['item_cnt_month']
X = train.drop(['item_cnt_month'], axis=1).values
#X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=1)
logger.debug('make_train_dat end')
logger.info('Fitting start')
forest = RandomForestRegressor(n_estimators=50, random_state=1)
forest.fit(X, y)
logger.debug('Fitting end')
logger.info('Scoring start')
#logger.info('Accuracy on test set: {:.3f}'.format(forest.score(X_test, y_test)))
test_data = load_test_data()
test = test_data.drop(['ID'], axis=1).values
submission = load_submission()
submission['item_cnt_month'] = forest.predict(test).astype(
np.float16).clip(0., 20.)
submission.to_csv('./result_tmp/submit_180826_1st.csv',
encoding='utf-8-sig',
index=False)
logger.info('submission:\n{}'.format(submission.head()))
logger.debug('RandomForestRegressor end')
logger.debug('====================')
def forest_submit():
logger = mylogger()
logger.info('RandomForestRegressor start')
logger.debug('make_train_data start')
#train = pd.read_csv('./result_tmp/scaled_train.csv')
train = pd.read_csv('./result_tmp/scaled_train_DateBlockNum.csv')
#train = train[train['date_block_num']==33] #直近1ヶ月
train = train.loc[(30 < train['date_block_num']) &
(train['date_block_num'] <= 33)] #直近3m
y = train['item_cnt_month']
X = train.drop(['item_cnt_month', 'date_block_num'], axis=1).values
#X = train.drop(['item_cnt_month'], axis=1).values
#X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=1)
logger.debug('make_train_data end')
logger.info('Fitting start')
forest = RandomForestRegressor(n_estimators=50, random_state=1)
forest.fit(X, y)
logger.debug('Fitting end')
#EDAしたいとき
#fti = forest.feature_importances_
#print('Feature Importances:')
#for i, feature in enumerate(train.colunms):
# print('\t{0:10s}:{1:>.6f}'.format(feature, fti[i]))
logger.info('Scoring start')
#logger.info('Accuracy on test set: {:.3f}'.format(.score(X_test, y_test)))
test_data = load_test_data()
test = test_data.drop(['ID'], axis=1).values
submission = load_submission()
submission['item_cnt_month'] = forest.predict(test).astype(
np.float16).clip(0., 20.)
#submission.to_csv('./result_tmp/submit_180826_1st.csv', encoding='utf-8-sig', index=False)
submission.to_csv('./result_tmp/submit_180827_31-33.csv',
encoding='utf-8-sig',
index=False)
logger.info('submission:\n{}'.format(submission.head()))
logger.debug('RandomForestRegressor end')
logger.debug('====================')
forest = RandomForestRegressor(n_estimators=50, random_state=1)
forest.fit(X, y)
logger.debug('Fitting end')
#EDAしたいとき
#fti = forest.feature_importances_
#print('Feature Importances:')
#for i, feature in enumerate(train.colunms):
# print('\t{0:10s}:{1:>.6f}'.format(feature, fti[i]))
logger.info('Scoring start')
#logger.info('Accuracy on test set: {:.3f}'.format(.score(X_test, y_test)))
test_data = load_test_data()
test = test_data.drop(['ID'], axis=1).values
submission = load_submission()
submission['item_cnt_month'] = forest.predict(test).astype(np.float16).clip(0., 20.)
#submission.to_csv('./result_tmp/submit_180826_1st.csv', encoding='utf-8-sig', index=False)
submission.to_csv('./result_tmp/submit_180827_31-33.csv', encoding='utf-8-sig', index=False)
logger.info('submission:\n{}'.format(submission.head()))
logger.debug('RandomForestRegressor end')
logger.debug('====================')
# CV: KFold, StratifiedKFold, TimeSeriesSplit, GroupKFold
# [http://scikit-learn.org/stable/modules/cross_validation.html]
# TimeSeriesSplit()によるCV
# [http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.TimeSeriesSplit.html]
# 以下のイメージで分割する。TESTを予測するために、直近のTRAINを使うイメージがtscv
# TRAIN: [0] TEST: [1], TRAIN: [0 1] TEST: [2], TRAIN: [0 1 2] TEST: [3]
def forest_cv():
