def test(self):
"""Test the correctness by comparing ratios of the original histogram.
"""
counter = defaultdict(int)
size = 100000
sampler = RatingBasedSampler()
for s in itertools.islice(sampler(), size):
counter[s] += 1
ratios = [
abs(counter[normalize_rating(rating)] / size - value)
for rating, value in sampler._dist.items()
]
error = max(ratios) - min(ratios)
self.assertLess(error, size * 0.01)
def test_load_twice(self):
"""Test even if load function is called twice, it loads correct dataset.
"""
# Load a dataset.
g = loader.load(Graph(), self.input)
# Create another dataset which some nodes are overrapped.
max_reviewers = 15
max_products = 10
buf = StringIO.StringIO()
for r in range(5, max_reviewers):
for p in range(3, max_products):
member_id = "r{0}".format(r)
product_id = "p{0}".format(p)
if product_id in self.reviews[member_id]:
continue
rating = random.randint(1, 5)
self.reviews[member_id][
product_id] = normalize_rating(rating)
self.size += 1
json.dump({
"member_id": member_id,
"product_id": product_id,
"rating": rating,
"date": "2014-01-01"
}, buf)
buf.write("\n")
# Load another dataset.
g = loader.load(g, buf.getvalue().strip().split("\n"))
self.assertEqual(len(g.review), self.size)
for r in g.review:
self.assertIn(r.member_id, self.reviews)
self.assertIn(r.product_id, self.reviews[r.member_id])
self.assertAlmostEqual(
r.rating, self.reviews[r.member_id][r.product_id])
def setUp(self):
"""Create a sample graph and its JSON data.
"""
self.graph = Graph()
reviewers = {}
products = {}
for r in range(10):
for p in range(5):
if random.random() > 0.5:
continue
member_id = "r{0}".format(r)
product_id = "p{0}".format(p)
rating = normalize_rating(random.randint(1, 5))
if member_id not in reviewers:
reviewers[member_id] = self.graph.new_reviewer(member_id)
if product_id not in products:
products[product_id] = self.graph.new_product(product_id)
self.graph.add_review(reviewers[member_id],
products[product_id], rating,
"2014-01-01")
def test_print_state(self):
"""Test print_state with simple data.
"""
i = "abs"
output = StringIO.StringIO()
helper.print_state(self.graph, i, output)
for line in output.getvalue().split("\n"):
if not line:
continue
obj = json.loads(line)
self.assertEqual(obj["iteration"], i)
if "reviewer" in obj:
rid = obj["reviewer"]["reviewer_id"]
score = obj["reviewer"]["score"]
self.assertIn(rid, self.reviewers)
self.assertEqual(score, self.reviewers[rid].anomalous_score)
else:
pid = obj["product"]["product_id"]
summary = helper.normalize_rating(obj["product"]["summary"])
self.assertIn(pid, self.products)
self.assertAlmostEqual(summary, self.products[pid].summary)
def setUp(self):
"""Create a sample graph and its JSON data.
"""
self.reviews = defaultdict(dict)
self.size = 0
buf = StringIO.StringIO()
for r in range(10):
for p in range(5):
if random.random() > 0.5:
member_id = "r{0}".format(r)
product_id = "p{0}".format(p)
rating = random.randint(1, 5)
self.reviews[member_id][
product_id] = normalize_rating(rating)
self.size += 1
json.dump({
"member_id": member_id,
"product_id": product_id,
"rating": rating,
"date": "2014-01-01"
}, buf)
buf.write("\n")
self.input = buf.getvalue().split("\n")
def test(self):
"""Test with simple inputs.
"""
self.assertEqual(helper.normalize_rating(5), 1.)
self.assertEqual(helper.normalize_rating(1), 0.)
self.assertEqual(helper.normalize_rating(3), 0.5)