def run():
_user_ratings = int(request.args.get('user-ratings'))
_item_ratings = int(request.args.get('item-ratings'))
_splitter = request.args.get('splitter')
_ratio = float(request.args.get('ratio')) / 100
_k = int(request.args.get('length'))
loader = sequeval.UIRTLoader(user_ratings=_user_ratings,
item_ratings=_item_ratings)
ratings = loader.load('datasets/yes_reduced.csv')
builder = sequeval.Builder('1000 s')
sequences, items = builder.build(ratings)
profiler = sequeval.Profiler(sequences)
response = {
'profiler': {
'users': profiler.users(),
'items': profiler.items(),
'ratings': profiler.ratings(),
'sequences': profiler.sequences(),
'sparsity': parse(profiler.sparsity()),
'length': parse(profiler.sequence_length())
}
}
if _splitter == 'random':
splitter = sequeval.RandomSplitter(_ratio)
else:
splitter = sequeval.TimestampSplitter(_ratio)
training_set, test_set = splitter.split(sequences)
response['splitter'] = {
'training': len(training_set),
'test': len(test_set)
}
evaluator = sequeval.Evaluator(training_set, test_set, items, _k)
cosine = sequeval.CosineSimilarity(training_set, items)
response['evaluator'] = []
most_popular = baseline.MostPopularRecommender(training_set, items)
response['evaluator'].append(evaluation(evaluator, most_popular, cosine))
random = baseline.RandomRecommender(training_set, items)
response['evaluator'].append(evaluation(evaluator, random, cosine))
unigram = baseline.UnigramRecommender(training_set, items)
response['evaluator'].append(evaluation(evaluator, unigram, cosine))
bigram = baseline.BigramRecommender(training_set, items)
response['evaluator'].append(evaluation(evaluator, bigram, cosine))
return json.dumps(response)
print("\n# Parameters")
print("File:", args.file)
print("Seed:", args.seed)
print("User ratings:", args.user)
print("Item ratings:", args.item)
print("Delta:", args.delta)
print("Ratio:", args.ratio)
print("Length:", args.length)
# Set the random seed
if args.seed is not None:
random.seed(args.seed)
np.random.seed(args.seed)
loader = sequeval.UIRTLoader(user_ratings=args.user, item_ratings=args.item)
ratings = loader.load(args.file)
builder = sequeval.Builder(args.delta)
sequences, items = builder.build(ratings)
print("\n# Profiler")
profiler = sequeval.Profiler(sequences)
print("Users:", profiler.users())
print("Items:", profiler.items())
print("Ratings:", profiler.ratings())
print("Sequences:", profiler.sequences())
print("Sparsity:", profiler.sparsity())
print("Length:", profiler.sequence_length())
if args.splitter == 'random':
def test_uirt_threshold(self):
fake_input = "1,1,5,0\n1,2,5,5\n1,1,5,10\n1,3,5,30\n1,2,5,35,\n2,1,5,5\n2,1,5,30\n2,2,5,35\n2,2,2,40"
loader = sequeval.UIRTLoader(threshold=3)
self.assertEqual(ratings[0:8], loader.load(io.StringIO(fake_input)))
def test_uirt_skip(self):
fake_input = "1,1,5,0\n1,2,5,5\n1,1,5,10\n1,3,5,30\n1,2,5,35,\n2,1,5,5\n2,1,5,30\n2,2,5,35\n2,2,5,40"
loader = sequeval.UIRTLoader(skip=1)
self.assertEqual(ratings[1:], loader.load(io.StringIO(fake_input)))
def test_uirt_item_ratings(self):
fake_input = "1,1,5,0\n1,2,5,5\n1,1,5,10\n1,3,5,30\n1,2,5,35,\n2,1,5,5\n2,1,5,30\n2,2,5,35\n2,2,5,40"
loader = sequeval.UIRTLoader(item_ratings=2)
self.assertEqual(ratings[:3] + ratings[4:],
loader.load(io.StringIO(fake_input)))