def __init__(self, config, *args, **kwargs):
"""
Constructor of DataSet
:param path_train_data: relative path for train file
:param path_test_data: relative path for test file
"""
self.logger = logging.get_logger(self.__class__.__name__)
self.args = args
self.kwargs = kwargs
self.config = config
self.column_names = ['userId', 'itemId', 'rating', 'timestamp']
if config.config_test:
return
if config.data_config.strategy == "fixed":
path_train_data = config.data_config.train_path
path_val_data = getattr(config.data_config, "validation_path",
None)
path_test_data = config.data_config.test_path
visual_feature_path = getattr(config.data_config.side_information,
"visual_features", None)
item_mapping_path = getattr(config.data_config.side_information,
"item_mapping", None)
size_tuple = getattr(config.data_config.side_information,
"output_image_size", None)
if visual_feature_path and item_mapping_path:
feature_set = set(
pd.read_csv(item_mapping_path, sep="\t",
header=None)[0].unique().tolist())
else:
feature_set = {}
images_src_folder = getattr(config.data_config.side_information,
"images_src_folder", None)
if images_src_folder:
image_set = {
int(path.split(".")[0])
for path in os.listdir(images_src_folder)
}
else:
image_set = {}
if feature_set and image_set:
visual_set = feature_set and image_set
elif feature_set:
visual_set = feature_set
elif image_set:
visual_set = image_set
else:
visual_set = {}
self.side_information_data = SimpleNamespace()
self.train_dataframe, self.side_information_data.aligned_items = self.load_dataset_dataframe(
path_train_data, "\t", visual_set)
self.side_information_data.visual_feature_path = visual_feature_path
self.side_information_data.item_mapping_path = item_mapping_path
self.side_information_data.images_src_folder = images_src_folder
self.side_information_data.size_tuple = size_tuple
self.train_dataframe = self.check_timestamp(self.train_dataframe)
self.logger.info('{0} - Loaded'.format(path_train_data))
self.test_dataframe = pd.read_csv(path_test_data,
sep="\t",
header=None,
names=self.column_names)
self.test_dataframe = self.check_timestamp(self.test_dataframe)
if path_val_data:
self.validation_dataframe = pd.read_csv(
path_val_data,
sep="\t",
header=None,
names=self.column_names)
self.validation_dataframe = self.check_timestamp(
self.validation_dataframe)
self.tuple_list = [([
(self.train_dataframe, self.validation_dataframe)
], self.test_dataframe)]
else:
self.tuple_list = [(self.train_dataframe, self.test_dataframe)]
elif config.data_config.strategy == "hierarchy":
self.tuple_list = self.read_splitting(
config.data_config.root_folder)
elif config.data_config.strategy == "dataset":
self.logger.info("There will be the splitting")
path_dataset = config.data_config.dataset_path
visual_feature_path = getattr(config.data_config.side_information,
"visual_features", None)
item_mapping_path = getattr(config.data_config.side_information,
"item_mapping", None)
size_tuple = getattr(config.data_config.side_information,
"output_image_size", None)
if visual_feature_path and item_mapping_path:
feature_set = set(
pd.read_csv(item_mapping_path, sep="\t",
header=None)[0].unique().tolist())
else:
feature_set = {}
images_src_folder = getattr(config.data_config.side_information,
"images_src_folder", None)
if images_src_folder:
image_set = {
int(path.split(".")[0])
for path in os.listdir(images_src_folder)
}
else:
image_set = {}
if feature_set and image_set:
visual_set = feature_set and image_set
elif feature_set:
visual_set = feature_set
elif image_set:
visual_set = image_set
else:
visual_set = {}
self.side_information_data = SimpleNamespace()
self.dataframe, self.side_information_data.aligned_items = self.load_dataset_dataframe(
path_dataset, "\t", visual_set)
self.side_information_data.visual_feature_path = visual_feature_path
self.side_information_data.item_mapping_path = item_mapping_path
self.side_information_data.images_src_folder = images_src_folder
self.side_information_data.size_tuple = size_tuple
self.dataframe = self.check_timestamp(self.dataframe)
self.logger.info('{0} - Loaded'.format(path_dataset))
self.dataframe = PreFilter.filter(self.dataframe, self.config)
splitter = Splitter(self.dataframe, self.config.splitting)
self.tuple_list = splitter.process_splitting()
else:
raise Exception("Strategy option not recognized")
def __init__(self, config, *args, **kwargs):
"""
Constructor of DataSet
:param path_train_data: relative path for train file
:param path_test_data: relative path for test file
"""
self.logger = logging.get_logger(self.__class__.__name__)
self.args = args
self.kwargs = kwargs
self.config = config
self.column_names = ['userId', 'itemId', 'rating', 'timestamp']
if config.config_test:
return
if config.data_config.strategy == "fixed":
path_train_data = config.data_config.train_path
path_val_data = getattr(config.data_config, "validation_path",
None)
path_test_data = config.data_config.test_path
self.train_dataframe = pd.read_csv(path_train_data,
sep="\t",
header=None,
names=self.column_names)
self.train_dataframe = self.check_timestamp(self.train_dataframe)
self.logger.info(f"{path_train_data} - Loaded")
self.test_dataframe = pd.read_csv(path_test_data,
sep="\t",
header=None,
names=self.column_names)
self.test_dataframe = self.check_timestamp(self.test_dataframe)
if path_val_data:
self.validation_dataframe = pd.read_csv(
path_val_data,
sep="\t",
header=None,
names=self.column_names)
self.validation_dataframe = self.check_timestamp(
self.validation_dataframe)
self.tuple_list = [([
(self.train_dataframe, self.validation_dataframe)
], self.test_dataframe)]
else:
self.tuple_list = [(self.train_dataframe, self.test_dataframe)]
elif config.data_config.strategy == "hierarchy":
self.tuple_list = self.read_splitting(
config.data_config.root_folder)
elif config.data_config.strategy == "dataset":
self.logger.info("There will be the splitting")
path_dataset = config.data_config.dataset_path
self.dataframe = pd.read_csv(path_dataset,
sep="\t",
header=None,
names=self.column_names)
self.dataframe = self.check_timestamp(self.dataframe)
self.logger.info(('{0} - Loaded'.format(path_dataset)))
self.dataframe = PreFilter.filter(self.dataframe, self.config)
splitter = Splitter(self.dataframe, self.config.splitting)
self.tuple_list = splitter.process_splitting()
else:
raise Exception("Strategy option not recognized")
def __init__(self, config, *args, **kwargs):
"""
Constructor of DataSet
:param path_train_data: relative path for train file
:param path_test_data: relative path for test file
"""
self.logger = logging.get_logger(self.__class__.__name__)
self.args = args
self.kwargs = kwargs
self.config = config
self.column_names = ['userId', 'itemId', 'rating', 'timestamp']
if config.config_test:
return
self.side_information_data = SimpleNamespace()
if config.data_config.strategy == "fixed":
path_train_data = config.data_config.train_path
path_val_data = getattr(config.data_config, "validation_path",
None)
path_test_data = config.data_config.test_path
work_directory_path = config.data_config.side_information.work_directory
map_path = config.data_config.side_information.map
features_path = config.data_config.side_information.features
predicates_path = config.data_config.side_information.predicates
self.train_dataframe, self.side_information_data.feature_map, self.side_information_data.predicate_mapping = self.load_dataset_dataframe(
path_train_data, predicates_path, features_path)
self.train_dataframe = self.check_timestamp(self.train_dataframe)
self.logger.info(f"{path_train_data} - Loaded")
self.test_dataframe = pd.read_csv(path_test_data,
sep="\t",
header=None,
names=self.column_names)
self.test_dataframe = self.check_timestamp(self.test_dataframe)
if config.binarize == True:
self.test_dataframe["rating"] = 1
self.train_dataframe["rating"] = 1
if path_val_data:
self.validation_dataframe = pd.read_csv(
path_val_data,
sep="\t",
header=None,
names=self.column_names)
self.validation_dataframe = self.check_timestamp(
self.validation_dataframe)
self.tuple_list = [([
(self.train_dataframe, self.validation_dataframe)
], self.test_dataframe)]
else:
self.tuple_list = [(self.train_dataframe, self.test_dataframe)]
elif config.data_config.strategy == "hierarchy":
item_mapping_path = getattr(config.data_config.side_information,
"item_mapping", None)
self.side_information_data.feature_map = self.load_attribute_file(
item_mapping_path)
self.tuple_list = self.read_splitting(
config.data_config.root_folder)
self.logger.info('{0} - Loaded'.format(
config.data_config.root_folder))
elif config.data_config.strategy == "dataset":
self.logger.info("There will be the splitting")
path_dataset = config.data_config.dataset_path
work_directory_path = config.data_config.side_information.work_directory
map_path = config.data_config.side_information.map
features_path = config.data_config.side_information.features
predicates_path = config.data_config.side_information.predicates
self.dataframe, self.side_information_data.feature_map, self.side_information_data.predicate_mapping = self.load_dataset_dataframe(
path_dataset, predicates_path, features_path)
self.dataframe = self.check_timestamp(self.dataframe)
self.logger.info(('{0} - Loaded'.format(path_dataset)))
self.dataframe = PreFilter.filter(self.dataframe, self.config)
if config.binarize == True:
self.dataframe["rating"] = 1
splitter = Splitter(self.dataframe, self.config.splitting)
self.tuple_list = splitter.process_splitting()
else:
raise Exception("Strategy option not recognized")