def execute_evaluation_markov_all_users_vs_one(train_tags, test_tags, user_id, execution_id, input_data_version,
is_distributive, random_dummy_mode, dir_name, save_result):
test_data = test_markov(train_tags, test_tags, is_distributive=is_distributive, random_dummy_mode=random_dummy_mode)
test_data["algorithm"] = "markov"
test_data["trained_with"] = "all_other_users"
test_data["train_size"] = len(train_tags)
test_data["test_size"] = len(test_tags)
if random_dummy_mode is None:
test_data["is_dummy"] = True
else:
test_data["is_dummy"] = False
test_data["method"] = "all_users_vs_one"
test_data["user_id"] = user_id
test_data["is_distributive"] = is_distributive
test_data["input_data_version"] = input_data_version
test_data["test_id"] = execution_id
if save_result:
experiments_dao.save_execution_test_data(result_dict=test_data,
filename=dir_name + "/" + test_data["test_id"])
def evaluation_ctw_single_partition_light_mem(tags_sequence,
user_id,
input_data_version,
predict_choice_method,
dir_name,
depth,
repeats_n=3,
save_result=True):
if len(tags_sequence) <= 1:
print("sr_group size: {} \n skipping".format(len(tags_sequence)))
raise exceptions.TooShortStopRegionGroup()
execution_id = str(uuid.uuid4())
for repeat_i in range(repeats_n):
test_data = test_ctw(tags_sequence,
depth=depth,
predict_choice_method=predict_choice_method)
test_data["algorithm"] = "ctw"
test_data["trained_with"] = "same_user"
test_data["train_size"] = len(tags_sequence)
test_data["test_size"] = len(tags_sequence)
if predict_choice_method == "random_dummy":
test_data["is_dummy"] = True
else:
test_data["is_dummy"] = False
test_data["pred_choice_method"] = predict_choice_method
test_data["method"] = "single_partition"
test_data["k"] = None
test_data["iteration"] = repeat_i
test_data["user_id"] = user_id
test_data["is_distributive"] = False
test_data["input_data_version"] = input_data_version
test_data["test_id"] = execution_id
if save_result:
experiments_dao.save_execution_test_data(result_dict=test_data,
filename=dir_name + "/" +
test_data["test_id"] +
"_i_{}".format(repeat_i))
def evaluation_markov_cluster_light_mem(cluster, test_user_id, input_data_version, cluster_version,
dir_name, repeats_n=3, is_distributive=False,
random_dummy_mode=None, save_result=True):
execution_id = str(uuid.uuid4())
if len(cluster) <= 1:
raise exceptions.ClusterSizeInadequate
user_tags, rest_cluster = partition_dict_by_keys_one_vs_all(a_dict=cluster, split_key=test_user_id)
for repeat_i in range(repeats_n):
test_data = test_markov_cluster(train_cluster=[rest_cluster[user_id] for user_id in rest_cluster.keys()],
test=user_tags,
is_distributive=is_distributive,
random_dummy_mode=random_dummy_mode)
test_data["algorithm"] = "markov"
test_data["trained_with"] = "cluster"
test_data["train_size"] = sum([len(rest_cluster[train_user_id]) for train_user_id in rest_cluster.keys()])
test_data["test_size"] = len(user_tags)
if random_dummy_mode is None:
test_data["is_dummy"] = False
else:
test_data["is_dummy"] = True
test_data["method"] = "cluster"
test_data["cluster_size"] = len(cluster)
test_data["k"] = None
test_data["iteration"] = repeat_i
test_data["user_id"] = test_user_id
test_data["multi_trip"] = False
test_data["is_distributive"] = is_distributive
test_data["input_data_version"] = input_data_version
test_data["cluster_version"] = cluster_version
test_data["test_id"] = execution_id
if save_result:
experiments_dao.save_execution_test_data(result_dict=test_data,
filename=dir_name + "/" + test_data["test_id"] + "_i_{}".format(repeat_i))
def evaluation_markov_k_fold_light_mem(tags_sequence, user_id, input_data_version, dir_name, k=5, is_distributive=False, random_dummy_mode=None, save_result=True):
if len(tags_sequence) <= 1:
print("sr_group size: {} \n skipping".format(len(tags_sequence)))
raise exceptions.TooShortStopRegionGroup()
k_fold_partitions = k_fold_iteration(tags_sequence, k)
execution_id = str(uuid.uuid4())
for i in range(len(k_fold_partitions)):
partition = k_fold_partitions[i]
train = partition["train"]
test = partition["test"]
test_data = test_markov(train, test, is_distributive=is_distributive, random_dummy_mode=random_dummy_mode)
test_data["algorithm"] = "markov"
test_data["trained_with"] = "same_user"
test_data["train_size"] = len(train)
test_data["test_size"] = len(test)
if random_dummy_mode is None:
test_data["is_dummy"] = True
else:
test_data["is_dummy"] = False
test_data["method"] = "k_fold"
test_data["k"] = k
test_data["iteration"] = i
test_data["user_id"] = user_id
test_data["is_distributive"] = is_distributive
test_data["input_data_version"] = input_data_version
test_data["test_id"] = execution_id
if save_result:
experiments_dao.save_execution_test_data(result_dict=test_data, filename=dir_name + "/" + test_data["test_id"] + "_i_{}".format(i))