def init():
global weight_mode, sim_threshold, sim_mode, attack_ratings, original_ratings, user_size, item_size, translator, correct, total, eccen
parser = argparse.ArgumentParser(
description='a ns attack simulation and statitical analysis')
parser.add_argument('-r', '--ratings', required=True)
parser.add_argument('-t', '--total', required=True, type=int)
parser.add_argument('-c', '--correct', required=True, type=int)
parser.add_argument('-e', '--eccen', type=float)
parser.add_argument('-m,', '--mode', choices=mode_choices, default='exp')
parser.add_argument('-w',
'--weight',
choices=['equal', 'less'],
default='less')
args = parser.parse_args()
weight_mode = args.weight
sim_mode = args.mode
attack_ratings = load(args.ratings)
dataset = extract_dataset_from_filename(args.ratings)
original_ratings = load(get_ratings_name_from_dataset(dataset))
user_size = original_ratings.shape[0]
item_size = original_ratings.shape[1]
translator = get_id_translator(args.ratings)
correct = args.correct
total = args.total
eccen = args.eccen
def main():
parser = argparse.ArgumentParser(description='RMSE test for a certain dataset')
parser.add_argument('-d', '--dataset', required=True, choices=dataset_choices)
parser.add_argument('-w', '--web', required=True)
parser.add_argument('-t', '--top', type=int)
parser.add_argument('-a', '--adapter')
args = parser.parse_args()
data_set = args.dataset
web_name = args.web
top = args.top
adapter_kind = args.adapter
temp_tops = [5, 10, 20, 40, 60, 70, 80, 90, 100]
original_ratings = load(get_ratings_name_from_dataset(data_set))
test_set = np.loadtxt(directory + data_set + '.test', delimiter='\t')
if web_name == 'all' and adapter_kind != 'all':
web_names = get_all_web_files()
plt.figure()
plt.xlabel('top')
plt.ylabel('RMSE')
best = {'top': None, 'RMSE': 2, 'web': None}
for web_name in web_names:
y = []
web = load(web_name)
for temp_top in temp_tops:
logging.info({'web': web_name, 'top': temp_top, 'adapter': adapter_kind})
count = RMSE(test_set, original_ratings, web, temp_top, adapter_kind)
y.append(count['RMSE'])
if min(y) < best['RMSE']:
best = {'web': web_name, 'RMSE': min(y), 'top': temp_tops[y.index(min(y))]}
logging.info('%s:%s', web_name, y)
plt.plot(temp_tops, y, marker='*', label=web_name)
plt.legend()
plt.savefig('top-RMSE-webs.jpg')
logging.info('best:%s', best)
elif web_name != 'all' and adapter_kind == 'all':
adapter_kinds = ['int', 'round', 'customize', 'int1']
plt.figure()
plt.xlabel('top')
plt.ylabel('RMSE')
web = load(web_name)
best = {'top': None, 'RMSE': 2, 'adapter': None}
for adapter_kind in adapter_kinds:
y = []
for temp_top in temp_tops:
logging.info({'web': web_name, 'top': temp_top, 'adapter': adapter_kind})
count = RMSE(test_set, original_ratings, web, temp_top, adapter_kind)
y.append(count['RMSE'])
if min(y) < best['RMSE']:
best = {'adapter': adapter_kind, 'RMSE': min(y), 'top': temp_tops[y.index(min(y))]}
logging.info('%s:%s', adapter_kind, y)
plt.plot(temp_tops, y, marker='*', label=adapter_kind)
plt.legend()
plt.savefig('top-RMSE-adapters.jpg')
logging.info('best:%s', best)
elif web_name != 'all' and adapter_kind != 'all':
logging.info(RMSE(test_set, original_ratings, load(web_name), top, adapter_kind))
else:
raise ValueError
def main():
parser = argparse.ArgumentParser(
description='k corating a rating file by a certain web')
parser.add_argument('-d', '--database', required=True)
parser.add_argument('-k', type=int, required=True)
parser.add_argument('-m', '--mode', required=True)
parser.add_argument('-a', '--analysis', action='store_true')
parser.add_argument('-t', '--trial', type=int, default=default_trial_times)
args = parser.parse_args()
original_ratings = load(get_ratings_name_from_dataset(args.database))
trial_times = args.trial
k = args.k
mode = args.mode
need_analysis = args.analysis
if os.path.exists('nratings_' + args.database + '.csv'):
normalized_ratings = load('nratings_' + args.database)
else:
normalized_ratings = np.zeros(shape=(original_ratings.shape),
dtype=int)
for i in range(original_ratings.shape[0]):
normalized_ratings[i, :] = normalize(original_ratings[i, :])
dump('nratings_%s.csv' % args.database, normalized_ratings)
Cluster.normalized_ratings = normalized_ratings
if os.path.exists('dis_map.csv'):
dis_map = load('dis_map.csv')
else:
dis_map = None
if k != 0:
best_seeds = get_best_initial_seeds(original_ratings,
normalized_ratings, k, mode,
dis_map, trial_times)
best_clusters = k_means(best_seeds)
dump_clusters(best_clusters)
if need_analysis:
analysis_of_clusters(best_clusters)
else:
if mode == 'all':
modes = ['density', 'dsort', 'rsort', 'random']
else:
modes = [mode]
find_best_k(original_ratings, normalized_ratings, modes, dis_map,
trial_times)
def k_corated(webname):
global sorted_ratings
if cluster_flag:
clusters = load_clusters(load('nratings_' + data_set), k)
sorted_ratings = sort_through_clusters(original_ratings, clusters)
k_corated, index_translator = corating_all_through_clusters(
sorted_ratings.copy(), clusters)
else:
sorted_ratings = sort(original_ratings)
k_corated, index_translator = k_corating_all(
sorted_ratings.copy(), k)
k_file_name = get_k_corated_name_by_attr(data_set, k, webname)
index_file_name = get_k_corated_index_by_attr(data_set, k, webname)
dump(k_file_name, k_corated)
dump(index_file_name, index_translator)
if i != 0:
index = int(i) - 1
count[index] += 1
return count
all_count = [0] * rating_scale
items_count = []
item_size = original_ratings.shape[1]
for i in range(item_size):
item = original_ratings[:, i]
temp_count = get_count(item)
items_count.append(temp_count)
all_count = list(np.array(all_count) + np.array(temp_count))
# for i in range(rating_scale):
# plt.figure()
# label = 'portion of %s' % (i + 1)
# x = [i for i in range(item_size)]
# y = [temp_count[i] / sum(temp_count) for temp_count in items_count]
# plt.plot(x, y, label=label)
# plt.legend()
# plt.savefig('items_portion_%s.jpg' % (i + 1))
logging.info([i / sum(all_count) for i in all_count])
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('-r', '--ratings', required=True)
args = parser.parse_args()
original_ratings = load(args.ratings)
ratia_analysis(original_ratings)
def init(dataset, web_name):
global original_ratings, web
original_ratings = load(get_ratings_name_from_dataset(dataset))
web = load(web_name)