def move_to_second_valid(best_select=[], path='', rank=0, gain=0, key_list=[]):
logger = logger_func()
if len(best_select)==0:
try:
if path=='':
path = sys.argv[2]
except IndexError:
pass
best_select = pd.read_csv(path)
try:
select_list = sys.argv[3].split('_')
select_type = select_list[0]
select_num = np.int(select_list[1])
except IndexError:
pass
if select_type=='rank':
best_feature = best_select.query(f"rank<={select_num}")['feature'].values
elif select_type=='gain':
best_feature = best_select.query(f"importance>={select_num}")['feature'].values
try:
best_feature = [col for col in best_feature if col.count(sys.argv[4])]
except IndexError:
best_feature = [col for col in best_feature if col.count('')]
if len(best_feature)==0:
sys.exit()
path_list = glob.glob('../features/*.gz')
select_path = []
for feature in best_feature:
for path in path_list:
filename = re.search(r'/([^/.]*).gz', path).group(1)
if filename[:3]=='tra':
filename = filename[6:]
if filename[:3]=='tes':
filename = filename[5:]
# if path.count(feature) and feature not in ignore_list:
if feature==filename:
select_path.append(path)
move_path = list(set(path_list) - set(select_path))
for move in move_path:
try:
shutil.move(move, '../features/no_use/')
except FileNotFoundError:
logger.info(f'FileNotFoundError: {feature}')
except shutil.Error:
logger.info(f'Shutil Error: {feature}')
print(f'move to third_valid:{len(best_feature)}')
def move_to_use():
try:
path = sys.argv[2]
except IndexError:
path = ''
best_select = pd.read_csv(path)
best_feature = best_select['feature'].values
win_list = glob.glob(win_path + '*')
first_list = glob.glob('../features/1_first_valid/*')
second_list = glob.glob('../features/2_second_valid/*')
third_list = glob.glob('../features/3_third_valid/*')
tmp_list = glob.glob('../features/5_tmp/*')
path_list = third_list
# path_list = third_list + tmp_list + win_list
# path_list = first_list + second_list + third_list + tmp_list + win_list
done_list = []
for feature in best_feature:
for path in path_list:
try:
filename = re.search(r'/([^/.]*).gz', path).group(1)
except AttributeError:
continue
# if path.count(feature):
# if filename==feature:
if filename.replace('stan_', '')==feature:
try:
shutil.move(path, win_path)
# filename = re.search(r'/([^/.]*).gz', path).group(1)
done_list.append(filename)
except shutil.Error:
pass
# shutil.move(path, gdrive_path)
except FileNotFoundError:
pass
# shutil.move(path, gdrive_path)
logger = logger_func()
best_feature = [f for f in best_feature]
loss_list = set(list(best_feature)) - set(done_list)
logger.info(f"Loss List:")
for loss in loss_list:
logger.info(f"{loss}")
def move_to_second_valid(best_select=[], path='', rank=0, key_list=[]):
logger = logger_func()
if len(best_select) == 0:
try:
if path == '':
path = sys.argv[2]
except IndexError:
pass
best_select = pd.read_csv(path)
try:
if rank == 0:
rank = int(sys.argv[3])
except IndexError:
pass
best_feature = best_select.query(f"rank>={rank}")['feature'].values
try:
best_feature = [
col for col in best_feature if col.count(sys.argv[4])
]
except IndexError:
best_feature = [col for col in best_feature if col.count('')]
if len(best_feature) == 0:
sys.exit()
path_list = glob.glob('../features/4_winner/*')
for feature in best_feature:
move_path = []
for path in path_list:
filename = re.search(r'/([^/.]*).gz', path).group(1)
# if path.count(feature) and feature not in ignore_list:
# if feature==filename:
if feature == filename.replace('stan_', ''):
# print(f"{filename} | {feature}")
move_path.append(path)
for move in move_path:
try:
shutil.move(move, second_path)
except FileNotFoundError:
logger.info(f'FileNotFoundError: {feature}')
except shutil.Error:
logger.info(f'Shutil Error: {feature}')
print(f'move to third_valid:{len(best_feature)}')
def move_to_use():
try:
path = sys.argv[2]
except IndexError:
path = ''
best_select = pd.read_csv(path)
best_feature = best_select['feature'].values
win_list = glob.glob(win_path + '*')
first_list = glob.glob('../season1_features/1_first_valid/*')
second_list = glob.glob('../season1_features/2_second_valid/*')
third_list = glob.glob('../season1_features/3_third_valid/*')
tmp_list = glob.glob('../features/5_tmp/*')
path_list = third_list + tmp_list + win_list
# path_list = first_list + second_list + third_list + tmp_list + win_list
done_list = []
for feature in best_feature:
for path in path_list:
if path.replace('.0', '_0').count(feature[:7]) and path.replace(
'.0', '_0').count(feature[9:]):
try:
shutil.move(path, win_path)
filename = re.search(r'/([^/.]*).gz',
path.replace('.0', '_0')).group(1)
if filename.count('train'):
done_list.append(filename[14:])
elif filename.count('test'):
done_list.append(filename[13:])
except shutil.Error:
pass
# shutil.move(path, gdrive_path)
except FileNotFoundError:
pass
# shutil.move(path, gdrive_path)
logger = logger_func()
best_feature = [f[8:] for f in best_feature]
loss_list = set(list(best_feature)) - set(done_list)
logger.info(f"Loss List:")
for loss in loss_list:
logger.info(f"{loss}")
def move_to_second_valid(best_select=[], path='', rank=0, key_list=[]):
logger = logger_func()
if len(best_select) == 0:
try:
if path == '':
path = sys.argv[2]
except IndexError:
pass
best_select = pd.read_csv(path)
try:
if rank == 0:
rank = int(sys.argv[3])
except IndexError:
pass
best_feature = best_select.query(f"rank>={rank}")['feature'].values
try:
best_feature = [
col for col in best_feature if col.count(sys.argv[4])
]
except IndexError:
best_feature = [col for col in best_feature if col.count('')]
if len(best_feature) == 0:
sys.exit()
path_list = glob.glob('../season1_features/4_winner/*')
for feature in best_feature:
move_path = [
path for path in path_list if path.count(feature[:7])
and path.count(feature[9:]) and feature not in ignore_list
]
for move in move_path:
try:
shutil.move(move, second_path)
except FileNotFoundError:
print(f'FileNotFound. : {feature}.gz')
pass
except shutil.Error:
logger.info(f'Shutil Error: {feature}')
print(f'move to third_valid:{len(best_feature)}')
import numpy as np
import datetime
from sklearn.preprocessing import LabelEncoder
#========================================================================
# original library
HOME = os.path.expanduser('~')
sys.path.append(f"{HOME}/kaggle/data_analysis/library/")
sys.path.append(f"{HOME}/kaggle/data_analysis/model/")
import MS_utils
from params_MS import params_lgb
import utils, ml_utils
from utils import logger_func
try:
if not logger:
logger = logger_func()
except NameError:
logger = logger_func()
#========================================================================
"""
argv[1]: comment
argv[2]: feature_key
argv[3]: group
"""
# Columns
base_path = '../input/base_exclude*'
base_path = '../input/base_Av*'
key, target, ignore_list = MS_utils.get_basic_var()
ignore_list = [key, target, 'country_group', 'down_flg']
base = utils.read_df_pkl(base_path)[[key, target, 'country_group']]
