class PreferenceModelTest(unittest.TestCase):
def setUp(self):
self.data = ds.DataIO(verbose = False)
self.data.load('testdata/ratings.csv')
self.model = PreferenceModel(verbose = False)
self.model.build(self.data)
def test(self):
# test __extract_facts:
# overall number of ratings
self.assertTrue(self.model.P().sum() == len(self.data.ratings))
# item and user checksums
user_sums = list([0,]*self.data.num_users())
item_sums = list([0,]*self.data.num_items())
for (u,i,r) in self.data.ratings:
user_sums[u] += 1
item_sums[i] += 1
self.assertTrue(np.array_equal(self.model.P().sum(1), np.matrix(user_sums).T))
self.assertTrue(np.array_equal(self.model.P().sum(0), np.matrix(item_sums)))
def generate_solution(algo, n, given_items, filename):
# load the dataset
ds = dataset.DataIO(verbose=True)
ds.load('../data/ratings.csv')
given_items = ds.translate_items(given_items)
# build the recommender and generate predictions
rs = recsys.Recommender(ds,
PreferenceModel(verbose=True),
algo,
top_ns,
verbose=True)
rs.build()
recs = rs.recommend(given_items, n)
# save the predictions to file
output = ds.print_recs(recs,
given_items=given_items,
printer=coursera_pa1_printer)
file = open(filename, 'w')
file.write(output)
file.close()
print 'Predictions are written to %s' % filename
def setUp(self):
self.data = ds.DataIO(verbose = False)
self.data.load('testdata/ratings.csv')
self.model = PreferenceModel(verbose = False)
self.model.build(self.data)
