def setUp(self):
mock_broker = MagicMock()
self.recommender = SommelierRecommenderBase(b=mock_broker)
def setUp(self):
mock_broker = MagicMock()
self.recommender = SommelierRecommenderBase(b=mock_broker)
class RecommenderTest(unittest.TestCase):
dummy_tastings = [
{ "author_id": 1, "wine_id": 5, "rating": 9 },
{ "author_id": 2, "wine_id": 6, "rating": 10 },
{ "author_id": 3, "wine_id": 7, "rating": 11 },
{ "author_id": 4, "wine_id": 8, "rating": 12 }]
expected_preferences_1 = {
1: [ 0, 9, 0, 0 ],
2: [ 0, 0, 10, 0 ],
3: [ 0, 0, 0, 11 ],
4: [ 12, 0, 0, 0 ]
}, [8, 5, 6, 7]
expected_preferences_2 = {
5: [ 9, 0, 0, 0 ],
6: [ 0, 10, 0, 0 ],
7: [ 0, 0, 11, 0 ],
8: [ 0, 0, 0, 12 ]
}, [1, 2, 3, 4]
pearson_r_test_preferences = {
"row_1": [ 1, 2, 3, 4, 5 ],
"row_2": [ 2, 3, 4, 5, 6 ],
"row_3": [ 5, 4, 3, 2, 1 ],
"row_4": [ 3, 3, 3, 3, 3 ],
"row_5": [ 1, 2 ],
"row_6": [ 3, 1 ]}
dummy_line_tab_separated = "1\t2\t3\t4"
dummy_line_colon_separated = "1::2::3::4"
dummy_line_invalid_separator = "1BAZ2BAZ3BAZ4"
dummy_line_invalid_values = "Should::NOT::be:Strings"
expected_tasting = {'author_id': 1, 'rating': 3, 'tasting_date': '1970-01-01 01:00:04', 'wine_id': 2}
expected_movielens_lines_colon = [ "1::5::9::0", "2::6::10::0", "3::7::11::0", "4::8::12::0" ]
expected_movielens_lines_tab = [ "1\t5\t9\t0", "2\t6\t10\t0", "3\t7\t11\t0", "4\t8\t12\t0" ]
dummy_matrix_a = [[1,1,1,1],[2,2,2,2],[3,3,3,3]]
dummy_matrix_b = [[1,1,1,1],[2,4,4,2],[3,3,3,1]]
dummy_matrix_c = [[0,2,0,2],[0,2,0,2],[2,2,2,2]]
expected_mae_a_b = ((0.0,1.0,0.5),0.5)
expected_mae_a_c = ((1.0,1.0,1.0),1.0)
def setUp(self):
mock_broker = MagicMock()
self.recommender = SommelierRecommenderBase(b=mock_broker)
def test_pearson_r(self):
# rows 1 and 2 have 100% positive correlation
self.assertEqual(1.0, self.recommender.pearson_r(self.pearson_r_test_preferences, 'row_1', 'row_2'))
# rows 1 and 3 have 100% negative correlation
self.assertEqual(-1.0, self.recommender.pearson_r(self.pearson_r_test_preferences, 'row_1', 'row_3'))
# row 4 has a standard deviation of 0, which we need to deal with to avoid a division by zero error
# when the covariance is divided by the product of the standard deviations
# in this case the method returns 0.0, on the basis that a user submitting the same rating for every
# item is not expressing any preference at all, so a neutral similarity score is appropriate
self.assertEqual(0.0, self.recommender.pearson_r(self.pearson_r_test_preferences, 'row_1', 'row_4'))
# cases where there are two or less items for comparison are problematic for pearson_r; any two
# lists with 2 items will always result in either 1.0 or -1.0. That is not a useful score, as
# there may be no similarity in the ratings at all, so we return 0.0 if there are < 3 items
self.assertEqual(0.0, self.recommender.pearson_r(self.pearson_r_test_preferences, 'row_5', 'row_6'))
def test_preferences(self):
# preferences formatting for author rows / wine columns
self.assertEqual(self.expected_preferences_1, self.recommender.preferences(self.dummy_tastings, 'author_id', 'wine_id'))
# preferences formatting for wine rows / author columns
self.assertEqual(self.expected_preferences_2, self.recommender.preferences(self.dummy_tastings, 'wine_id', 'author_id'))
# Movielens data can be encoded with either tabs or double colons (::), so test for both and neither...
def test_movielens_line_to_tasting(self):
self.assertEqual(self.expected_tasting, self.recommender.movielens_line_to_tasting(self.dummy_line_tab_separated))
self.assertEqual(self.expected_tasting, self.recommender.movielens_line_to_tasting(self.dummy_line_colon_separated))
self.assertRaises(Exception, lambda _: self.recommender.movielens_line_to_tasting(self.dummy_line_invalid_separator))
self.assertRaises(Exception, lambda _: self.recommender.movielens_line_to_tasting(self.dummy_line_invalid_values))
# tastings_to_movielens_format() should use double colon (::) separator by default
def test_tastings_to_movielens_format(self):
self.assertEqual(self.expected_movielens_lines_colon, self.recommender.tastings_to_movielens_format(self.dummy_tastings))
self.assertEqual(self.expected_movielens_lines_colon, self.recommender.tastings_to_movielens_format(self.dummy_tastings, separator="::"))
self.assertEqual(self.expected_movielens_lines_tab, self.recommender.tastings_to_movielens_format(self.dummy_tastings, separator="\t"))
def test_evaluate_matrices(self):
self.assertEquals(self.expected_mae_a_b, self.recommender.recsys_evaluate_matrices(self.dummy_matrix_a, self.dummy_matrix_b))
self.assertEquals(self.expected_mae_a_c, self.recommender.recsys_evaluate_matrices(self.dummy_matrix_a, self.dummy_matrix_c))
class RecommenderTest(unittest.TestCase):
dummy_tastings = [{
"author_id": 1,
"wine_id": 5,
"rating": 9
}, {
"author_id": 2,
"wine_id": 6,
"rating": 10
}, {
"author_id": 3,
"wine_id": 7,
"rating": 11
}, {
"author_id": 4,
"wine_id": 8,
"rating": 12
}]
expected_preferences_1 = {
1: [0, 9, 0, 0],
2: [0, 0, 10, 0],
3: [0, 0, 0, 11],
4: [12, 0, 0, 0]
}, [8, 5, 6, 7]
expected_preferences_2 = {
5: [9, 0, 0, 0],
6: [0, 10, 0, 0],
7: [0, 0, 11, 0],
8: [0, 0, 0, 12]
}, [1, 2, 3, 4]
pearson_r_test_preferences = {
"row_1": [1, 2, 3, 4, 5],
"row_2": [2, 3, 4, 5, 6],
"row_3": [5, 4, 3, 2, 1],
"row_4": [3, 3, 3, 3, 3],
"row_5": [1, 2],
"row_6": [3, 1]
}
dummy_line_tab_separated = "1\t2\t3\t4"
dummy_line_colon_separated = "1::2::3::4"
dummy_line_invalid_separator = "1BAZ2BAZ3BAZ4"
dummy_line_invalid_values = "Should::NOT::be:Strings"
expected_tasting = {
'author_id': 1,
'rating': 3,
'tasting_date': '1970-01-01 01:00:04',
'wine_id': 2
}
expected_movielens_lines_colon = [
"1::5::9::0", "2::6::10::0", "3::7::11::0", "4::8::12::0"
]
expected_movielens_lines_tab = [
"1\t5\t9\t0", "2\t6\t10\t0", "3\t7\t11\t0", "4\t8\t12\t0"
]
dummy_matrix_a = [[1, 1, 1, 1], [2, 2, 2, 2], [3, 3, 3, 3]]
dummy_matrix_b = [[1, 1, 1, 1], [2, 4, 4, 2], [3, 3, 3, 1]]
dummy_matrix_c = [[0, 2, 0, 2], [0, 2, 0, 2], [2, 2, 2, 2]]
expected_mae_a_b = ((0.0, 1.0, 0.5), 0.5)
expected_mae_a_c = ((1.0, 1.0, 1.0), 1.0)
def setUp(self):
mock_broker = MagicMock()
self.recommender = SommelierRecommenderBase(b=mock_broker)
def test_pearson_r(self):
# rows 1 and 2 have 100% positive correlation
self.assertEqual(
1.0,
self.recommender.pearson_r(self.pearson_r_test_preferences,
'row_1', 'row_2'))
# rows 1 and 3 have 100% negative correlation
self.assertEqual(
-1.0,
self.recommender.pearson_r(self.pearson_r_test_preferences,
'row_1', 'row_3'))
# row 4 has a standard deviation of 0, which we need to deal with to avoid a division by zero error
# when the covariance is divided by the product of the standard deviations
# in this case the method returns 0.0, on the basis that a user submitting the same rating for every
# item is not expressing any preference at all, so a neutral similarity score is appropriate
self.assertEqual(
0.0,
self.recommender.pearson_r(self.pearson_r_test_preferences,
'row_1', 'row_4'))
# cases where there are two or less items for comparison are problematic for pearson_r; any two
# lists with 2 items will always result in either 1.0 or -1.0. That is not a useful score, as
# there may be no similarity in the ratings at all, so we return 0.0 if there are < 3 items
self.assertEqual(
0.0,
self.recommender.pearson_r(self.pearson_r_test_preferences,
'row_5', 'row_6'))
def test_preferences(self):
# preferences formatting for author rows / wine columns
self.assertEqual(
self.expected_preferences_1,
self.recommender.preferences(self.dummy_tastings, 'author_id',
'wine_id'))
# preferences formatting for wine rows / author columns
self.assertEqual(
self.expected_preferences_2,
self.recommender.preferences(self.dummy_tastings, 'wine_id',
'author_id'))
# Movielens data can be encoded with either tabs or double colons (::), so test for both and neither...
def test_movielens_line_to_tasting(self):
self.assertEqual(
self.expected_tasting,
self.recommender.movielens_line_to_tasting(
self.dummy_line_tab_separated))
self.assertEqual(
self.expected_tasting,
self.recommender.movielens_line_to_tasting(
self.dummy_line_colon_separated))
self.assertRaises(
Exception, lambda _: self.recommender.movielens_line_to_tasting(
self.dummy_line_invalid_separator))
self.assertRaises(
Exception, lambda _: self.recommender.movielens_line_to_tasting(
self.dummy_line_invalid_values))
# tastings_to_movielens_format() should use double colon (::) separator by default
def test_tastings_to_movielens_format(self):
self.assertEqual(
self.expected_movielens_lines_colon,
self.recommender.tastings_to_movielens_format(self.dummy_tastings))
self.assertEqual(
self.expected_movielens_lines_colon,
self.recommender.tastings_to_movielens_format(self.dummy_tastings,
separator="::"))
self.assertEqual(
self.expected_movielens_lines_tab,
self.recommender.tastings_to_movielens_format(self.dummy_tastings,
separator="\t"))
def test_evaluate_matrices(self):
self.assertEquals(
self.expected_mae_a_b,
self.recommender.recsys_evaluate_matrices(self.dummy_matrix_a,
self.dummy_matrix_b))
self.assertEquals(
self.expected_mae_a_c,
self.recommender.recsys_evaluate_matrices(self.dummy_matrix_a,
self.dummy_matrix_c))