def main():
# Data loading
root_data_path = os.path.join(get_project_root_path(), "data/")
data_reader = RecSys2019Reader(root_data_path)
data_reader = New_DataSplitter_leave_k_out(
data_reader,
k_out_value=K_OUT,
use_validation_set=False,
allow_cold_users=ALLOW_COLD_USERS,
force_new_split=True,
seed=get_split_seed())
data_reader.load_data()
URM_train, URM_test = data_reader.get_holdout_split()
ICM_all, _ = get_ICM_train_new(data_reader)
UCM_all, _ = get_UCM_train_new(data_reader)
# Ignoring users
ignore_users = get_ignore_users(
URM_train,
data_reader.get_original_user_id_to_index_mapper(),
lower_threshold=LOWER_THRESHOLD,
upper_threshold=UPPER_THRESHOLD,
ignore_non_target_users=IGNORE_NON_TARGET_USERS)
evaluator = EvaluatorHoldout(URM_test,
cutoff_list=[CUTOFF],
ignore_users=ignore_users)
# Model evaluation
model = get_model(URM_train, ICM_all, UCM_all)
print(evaluator.evaluateRecommender(model))
def main():
args = get_arguments()
# Data loading
root_data_path = args.reader_path
data_reader = RecSys2019Reader(root_data_path)
data_reader = New_DataSplitter_leave_k_out(data_reader, k_out_value=K_OUT, allow_cold_users=ALLOW_COLD_USERS,
use_validation_set=False, force_new_split=True, seed=args.seed)
data_reader.load_data()
URM_train, URM_test = data_reader.get_holdout_split()
# Remove interactions to users that has len == 1 to URM_train
len_1_users_mask = np.ediff1d(URM_train.tocsr().indptr) == 1
len_1_users = np.arange(URM_train.shape[0])[len_1_users_mask]
URM_train = URM_train.tolil()
URM_train[len_1_users, :] = 0
URM_train = URM_train.tocsr()
# Remove interactions to users that has len == 1 to URM_test
len_1_users_mask = np.ediff1d(URM_test.tocsr().indptr) == 1
len_1_users = np.arange(URM_test.shape[0])[len_1_users_mask]
URM_test = URM_test.tolil()
URM_test[len_1_users, :] = 0
URM_test = URM_test.tocsr()
UCM_all = get_UCM_train_cold(data_reader)
ignore_users = get_ignore_users(URM_train, data_reader.get_original_user_id_to_index_mapper(),
lower_threshold=args.lower_threshold, upper_threshold=args.upper_threshold,
ignore_non_target_users=args.exclude_non_target)
ignore_users = np.concatenate([ignore_users, len_1_users])
# Setting evaluator
cutoff_list = [10]
evaluator = EvaluatorHoldout(URM_test, cutoff_list=cutoff_list, ignore_users=ignore_users)
# HP tuning
print("Start tuning...")
version_path = "../../report/hp_tuning/{}/".format(args.recommender_name)
now = datetime.now().strftime('%b%d_%H-%M-%S')
now = now + "_k_out_value_{}/".format(K_OUT)
version_path = version_path + "/" + now
run_parameter_search_user_demographic(URM_train=URM_train, UCM_object=UCM_all, UCM_name="UCM_all",
recommender_class=RECOMMENDER_CLASS_DICT[args.recommender_name],
evaluator_validation=evaluator,
metric_to_optimize="MAP",
output_folder_path=version_path,
parallelizeKNN=True,
n_cases=int(args.n_cases),
n_random_starts=int(args.n_random_starts))
print("...tuning ended")
def main():
args = get_arguments()
# Data loading
data_reader = read_split_load_data(3, args.allow_cold_users, args.seed)
URM_train, URM_test = data_reader.get_holdout_split()
ICM_categorical = data_reader.get_ICM_from_name("ICM_sub_class")
ICM_numerical, _ = get_ICM_numerical(data_reader.dataReader_object)
ICM_all, _ = get_ICM_train_new(data_reader)
similarity_type_list = None
if args.recommender_name == "item_cbf_numerical":
ICM = ICM_numerical
ICM_name = "ICM_numerical"
elif args.recommender_name == "item_cbf_categorical":
ICM = ICM_categorical
ICM_name = "ICM_categorical"
else:
ICM = ICM_all
ICM_name = "ICM_all"
# Setting evaluator
ignore_users = get_ignore_users(
URM_train,
data_reader.get_original_user_id_to_index_mapper(),
lower_threshold=args.lower_threshold,
upper_threshold=args.upper_threshold,
ignore_non_target_users=args.exclude_non_target)
evaluator = EvaluatorHoldout(URM_test,
cutoff_list=[10],
ignore_users=ignore_users)
# HP tuning
print("Start tuning...")
version_path = "../../report/hp_tuning/{}/".format(args.recommender_name)
now = datetime.now().strftime('%b%d_%H-%M-%S')
now = now + "_k_out_value_3/"
version_path = version_path + "/" + now
run_parameter_search_item_content(
URM_train=URM_train,
ICM_object=ICM,
ICM_name=ICM_name,
recommender_class=RECOMMENDER_CLASS_DICT[args.recommender_name],
evaluator_validation=evaluator,
metric_to_optimize="MAP",
output_folder_path=version_path,
similarity_type_list=similarity_type_list,
parallelizeKNN=True,
n_cases=args.n_cases,
n_random_starts=args.n_random_starts)
print("...tuning ended")
UCM_all = get_UCM_train(data_reader)
model = HybridWeightedAverageRecommender(URM_train, normalize=NORMALIZE)
all_models = _get_all_models(URM_train=URM_train,
UCM_all=UCM_all,
ICM_all=ICM_all)
for model_name, model_object in all_models.items():
model.add_fitted_model(model_name, model_object)
print("The models added in the hybrid are: {}".format(
list(all_models.keys())))
# Setting evaluator
ignore_users = get_ignore_users(
URM_train,
data_reader.get_original_user_id_to_index_mapper(),
lower_threshold=LOWER_THRESHOLD,
upper_threshold=UPPER_THRESHOLD,
ignore_non_target_users=IGNORE_NON_TARGET_USERS)
evaluator = EvaluatorHoldout(URM_test,
cutoff_list=[CUTOFF],
ignore_users=ignore_users)
version_path = "../../report/hp_tuning/hybrid_weighted_avg"
now = datetime.now().strftime('%b%d_%H-%M-%S')
now = now + "_k_out_value_{}/".format(K_OUT)
version_path = version_path + "/" + now
run_parameter_search_hybrid(model,
metric_to_optimize="MAP",
evaluator_validation=evaluator,
train_df = preprocess_dataframe_after_reading(train_df)
y_train = train_df['label'].values + 1
train_df = train_df.drop(columns=["label"], inplace=False)
valid_df = preprocess_dataframe_after_reading(valid_df)
valid_df = valid_df.drop(columns=[], inplace=False)
_, non_zero_count, total = get_label_array(data_frame=train_df,
URM_train=URM_train)
y_valid, _, _ = get_label_array(data_frame=valid_df, URM_train=URM_test)
# Setting evaluator
mapper = data_reader.get_original_user_id_to_index_mapper()
ignore_users = get_ignore_users(URM_train,
mapper,
lower_threshold=20,
upper_threshold=2**16 - 1,
ignore_non_target_users=True)
evaluator = EvaluatorHoldout(URM_test,
cutoff_list=[10],
ignore_users=ignore_users)
total_users = np.arange(URM_train.shape[0])
mask = np.in1d(total_users, ignore_users, invert=True)
users_to_validate = total_users[mask]
# HP tuning
print("Start tuning...")
version_path = "../../report/hp_tuning/light_gbm/"
now = datetime.now().strftime('%b%d_%H-%M-%S')
now = now + "_k_out_value_3_eval/"
version_path = version_path + now
def main():
set_env_variables()
args = get_arguments()
seeds = get_seed_lists(args.n_folds, get_split_seed())
# --------- DATA LOADING SECTION --------- #
URM_train_list = []
ICM_train_list = []
UCM_train_list = []
evaluator_list = []
for fold_idx in range(args.n_folds):
# Read and split data
data_reader = read_split_load_data(K_OUT, args.allow_cold_users, seeds[fold_idx])
URM_train, URM_test = data_reader.get_holdout_split()
ICM_train, item_feature2range = get_ICM_train_new(data_reader)
UCM_train, user_feature2range = get_UCM_train_new(data_reader)
# Ignore users and setting evaluator
ignore_users = get_ignore_users(URM_train, data_reader.get_original_user_id_to_index_mapper(),
args.lower_threshold, args.upper_threshold,
ignore_non_target_users=args.exclude_non_target)
# Ignore users by age
# UCM_age = data_reader.get_UCM_from_name("UCM_age")
# age_feature_to_id_mapper = data_reader.dataReader_object.get_UCM_feature_to_index_mapper_from_name("UCM_age")
# age_demographic = get_user_demographic(UCM_age, age_feature_to_id_mapper, binned=True)
# ignore_users = np.unique(np.concatenate((ignore_users, get_ignore_users_age(age_demographic, AGE_TO_KEEP))))
URM_train_list.append(URM_train)
ICM_train_list.append(ICM_train)
UCM_train_list.append(UCM_train)
evaluator = EvaluatorHoldout(URM_test, cutoff_list=[CUTOFF], ignore_users=np.unique(ignore_users))
evaluator_list.append(evaluator)
# --------- HYPER PARAMETERS TUNING SECTION --------- #
print("Start tuning...")
hp_tuning_path = "../../../report/hp_tuning/" + args.recommender_name + "/"
date_string = datetime.now().strftime('%b%d_%H-%M-%S_k1_lt_{}/'.format(args.lower_threshold))
output_folder_path = hp_tuning_path + date_string
if args.recommender_name in COLLABORATIVE_RECOMMENDER_CLASS_DICT.keys():
run_cv_parameter_search(URM_train_list=URM_train_list,
recommender_class=RECOMMENDER_CLASS_DICT[args.recommender_name],
evaluator_validation_list=evaluator_list,
metric_to_optimize="MAP", output_folder_path=output_folder_path,
parallelize_search=args.parallelize, n_jobs=args.n_jobs,
n_cases=args.n_cases, n_random_starts=args.n_random_starts)
elif args.recommender_name in CONTENT_RECOMMENDER_CLASS_DICT.keys():
run_cv_parameter_search(URM_train_list=URM_train_list, ICM_train_list=ICM_train_list, ICM_name="ICM_all",
recommender_class=RECOMMENDER_CLASS_DICT[args.recommender_name],
evaluator_validation_list=evaluator_list,
metric_to_optimize="MAP", output_folder_path=output_folder_path,
parallelize_search=args.parallelize, n_jobs=args.n_jobs,
n_cases=args.n_cases, n_random_starts=args.n_random_starts)
elif args.recommender_name in DEMOGRAPHIC_RECOMMENDER_CLASS_DICT.keys():
run_cv_parameter_search(URM_train_list=URM_train_list, UCM_train_list=UCM_train_list, UCM_name="UCM_all",
recommender_class=RECOMMENDER_CLASS_DICT[args.recommender_name],
evaluator_validation_list=evaluator_list,
metric_to_optimize="MAP", output_folder_path=output_folder_path,
parallelize_search=args.parallelize, n_jobs=args.n_jobs,
n_cases=args.n_cases, n_random_starts=args.n_random_starts)
elif args.recommender_name in SIDE_INFO_CLASS_DICT:
temp_list = []
for i, URM in enumerate(URM_train_list):
temp = sps.vstack([URM, ICM_train_list[i].T], format="csr")
#temp = TF_IDF(temp).tocsr()
temp_list.append(temp)
run_cv_parameter_search(URM_train_list=temp_list,
recommender_class=RECOMMENDER_CLASS_DICT[args.recommender_name],
evaluator_validation_list=evaluator_list, metric_to_optimize="MAP",
output_folder_path=output_folder_path, parallelize_search=args.parallelize,
n_jobs=args.n_jobs, n_cases=args.n_cases, n_random_starts=args.n_random_starts)
print("...tuning ended")
data_reader = New_DataSplitter_leave_k_out(data_reader,
k_out_value=1,
allow_cold_users=False,
use_validation_set=False,
force_new_split=True,
seed=get_split_seed())
data_reader.load_data()
URM_train, URM_test = data_reader.get_holdout_split()
ICM_all, _ = get_ICM_train_new(data_reader)
UCM_all = get_UCM_train(data_reader)
ignore_users = get_ignore_users(
URM_train,
data_reader.get_original_user_id_to_index_mapper(),
lower_threshold=-1,
upper_threshold=22,
ignore_non_target_users=True)
# Setting evaluator
cutoff_list = [10]
evaluator = EvaluatorHoldout(URM_test,
cutoff_list=cutoff_list,
ignore_users=ignore_users)
version_path = "../../report/hp_tuning/bagging/"
now = datetime.now().strftime('%b%d_%H-%M-%S')
now = now + "_k_out_value_3/"
version_path = version_path + "/" + now
hyper_parameters_range = {}
sub_6 = best_models.ItemCF.get_model(URM_train=URM_train, load_model=True, save_model=True)
sub_6.RECOMMENDER_NAME = "ItemCF"
sub_7 = best_models.SLIM_BPR.get_model(URM_train=URM_train, load_model=True, save_model=True)
sub_list = [sub_0, sub_1, sub_2, sub_3, sub_4, sub_5, sub_6, sub_7]
pure_svd_param = {'num_factors': 50, 'n_oversamples': 3, 'n_iter': 20, 'feature_weighting': 'TF-IDF'}
pure_svd = NewPureSVDRecommender(URM_train)
pure_svd.fit(**pure_svd_param)
user_factors = np.array(pure_svd.USER_factors)
item_factors = np.array(pure_svd.ITEM_factors)
mapper = data_reader.get_original_user_id_to_index_mapper()
ignore_users = get_ignore_users(URM_train, mapper, lower_threshold=LOWER_THRESHOLD, upper_threshold=2 ** 16 - 1,
ignore_non_target_users=IGNORE_NON_TARGET_USERS)
main_recommender = main_rec
total_users = np.arange(URM_train.shape[0])
mask = np.in1d(total_users, ignore_users, invert=True)
user_to_validate = total_users[mask]
data_path = "../../data/"
# Retrieve data for boosting
train_df = get_train_dataframe_proportion(user_id_array=user_to_validate,
cutoff=TRAIN_CUTOFF,
main_recommender=main_recommender,
recommender_list=sub_list,
mapper=mapper,
URM_train=URM_train,
user_factors=user_factors,