def main():
parser = argparse.ArgumentParser()
parser.add_argument('train_pattern', type=str, help='pattern of names for training files')
parser.add_argument('test_pattern', type=str, help='pattern of names for testing files')
args = parser.parse_args()
train_files = glob.glob(args.train_pattern)
test_files = glob.glob(args.test_pattern)
if len(train_files) != len(test_files):
raise ValueError('Number of training files not equal to number of testing files')
popular = PopularityStrategy()
entropy = EntropyStrategy()
entropy_zero = EntropyZeroStrategy()
nums = [5, 10, 15]
for count, (train_n, test_n) in enumerate(zip(train_files, test_files)):
print "Processing %d fold with cold start" % count
train = pd.read_csv(train_n, header=None, names=['user', 'item', 'rating', 'time'])
test = pd.read_csv(test_n, header=None, names=['user', 'item', 'rating', 'time'])
test_ids = test.user.unique()
user_fit = train.user.isin(test_ids)
train_other_folds = train[~user_fit]
train_this_fold = train[user_fit]
for n in nums:
select_fn = 'rated'
list_fn = 'list'
popular.write_train_test_movielens(train_n, test_n, select_fn, list_fn,
train_this_fold, train_other_folds, test, n)
entropy.write_train_test_movielens(train_n, test_n, select_fn, list_fn,
train_this_fold, train_other_folds, test, n)
entropy_zero.write_train_test_movielens(train_n, test_n, select_fn, list_fn,
train_this_fold, train_other_folds, test, n)
def test_entropy_zero(self):
count1 = pd.Series(data=[30, 30, 30])
assert abs(EntropyZeroStrategy.entropy_zero(count1, 5, 100) - 0.21794) < 0.0001
count1 = pd.Series(data=[50])
assert abs(EntropyZeroStrategy.entropy_zero(count1, 5, 100) - 0.094520) < 0.0001
