def test_class_init_with_data_nonstring(self):
self.g = NaiveBayes([(1, 2), (1, 3)])
six.assertCountEqual(self, self.g.nodes(), [1, 2, 3])
six.assertCountEqual(self, self.g.edges(), [(1, 2), (1, 3)])
self.assertEqual(self.g.parent_node, 1)
self.assertSetEqual(self.g.children_nodes, {2, 3})
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (2, 3)])
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (3, 2)])
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (3, 4)])
def test_class_init_with_data_string(self):
self.g = NaiveBayes([('a', 'b'), ('a', 'c')])
six.assertCountEqual(self, self.g.nodes(), ['a', 'b', 'c'])
six.assertCountEqual(self, self.g.edges(), [('a', 'b'), ('a', 'c')])
self.assertEqual(self.g.parent_node, 'a')
self.assertSetEqual(self.g.children_nodes, {'b', 'c'})
self.assertRaises(ValueError, NaiveBayes, [('a', 'b'), ('b', 'c')])
self.assertRaises(ValueError, NaiveBayes, [('a', 'b'), ('c', 'b')])
self.assertRaises(ValueError, NaiveBayes, [('a', 'b'), ('d', 'e')])
def test_class_init_with_data_string(self):
self.g = NaiveBayes([("a", "b"), ("a", "c")])
six.assertCountEqual(self, list(self.g.nodes()), ["a", "b", "c"])
six.assertCountEqual(self, list(self.g.edges()), [("a", "b"), ("a", "c")])
self.assertEqual(self.g.parent_node, "a")
self.assertSetEqual(self.g.children_nodes, {"b", "c"})
self.assertRaises(ValueError, NaiveBayes, [("a", "b"), ("b", "c")])
self.assertRaises(ValueError, NaiveBayes, [("a", "b"), ("c", "b")])
self.assertRaises(ValueError, NaiveBayes, [("a", "b"), ("d", "e")])
class TestNaiveBayesMethods(unittest.TestCase):
def setUp(self):
self.G1 = NaiveBayes([('a', 'b'), ('a', 'c'),
('a', 'd'), ('a', 'e')])
self.G2 = NaiveBayes([('d', 'g'), ('d', 'l'),
('d', 's')])
def test_local_independencies(self):
self.assertEqual(self.G1.local_independencies('a'), Independencies())
self.assertEqual(self.G1.local_independencies('b'), Independencies(['b', ['e', 'c', 'd'], 'a']))
self.assertEqual(self.G1.local_independencies('c'), Independencies(['c', ['e', 'b', 'd'], 'a']))
self.assertEqual(self.G1.local_independencies('d'), Independencies(['d', ['b', 'c', 'e'], 'a']))
def test_active_trail_nodes(self):
self.assertListEqual(sorted(self.G2.active_trail_nodes('d')), ['d', 'g', 'l', 's'])
self.assertListEqual(sorted(self.G2.active_trail_nodes('g')), ['d', 'g', 'l', 's'])
self.assertListEqual(sorted(self.G2.active_trail_nodes('l')), ['d', 'g', 'l', 's'])
self.assertListEqual(sorted(self.G2.active_trail_nodes('s')), ['d', 'g', 'l', 's'])
def test_active_trail_nodes_args(self):
self.assertListEqual(sorted(self.G2.active_trail_nodes('d', observed='g')), ['d', 'l', 's'])
self.assertListEqual(sorted(self.G2.active_trail_nodes('l', observed='g')), ['d', 'l', 's'])
self.assertListEqual(sorted(self.G2.active_trail_nodes('s', observed=['g', 'l'])), ['d', 's'])
self.assertListEqual(sorted(self.G2.active_trail_nodes('s', observed=['d', 'l'])), ['s'])
def test_get_ancestors_of(self):
self.assertListEqual(sorted(self.G1._get_ancestors_of('b')), ['a', 'b'])
self.assertListEqual(sorted(self.G1._get_ancestors_of('e')), ['a', 'e'])
self.assertListEqual(sorted(self.G1._get_ancestors_of('a')), ['a'])
self.assertListEqual(sorted(self.G1._get_ancestors_of(['b', 'e'])), ['a', 'b', 'e'])
def tearDown(self):
del self.G1
del self.G2
def reset(self) -> PGMNaiveBayes:
'''Totally reset the Classifier'''
self.categories = {}
self.tokens = {}
self.cardinality = 1
self.total_documents = 0
self.total_tokens = 0
self.cpd_class = None
self.model = NaiveBayes()
self.model.add_node(Data.CATEGORY_NAME)
return self
def Bayesian_Net_Model(data):
cols = data.columns.values
n_cols = len(data.columns.values)
BN_Model = NaiveBayes()
BN_Model.add_nodes_from(cols)
edges = []
for i in cols:
if (i != "Overall"):
edge = ["Overall", i]
edges.append(edge)
BN_Model.add_edges_from(edges)
print("Aggiunti Archi e Nodi \n\n")
data_cpts = Compute_CPT(data, "Overall")
CPTS_list = Generate_CPTs(data_cpts, data, cols, 'Overall')
test_list = [None] * len(CPTS_list)
for i in CPTS_list:
BN_Model.add_cpds(i)
print("Aggiunte CPD \n\n")
return BN_Model
def test_class_init_with_data_string(self):
self.g = NaiveBayes([('a', 'b'), ('a', 'c')])
six.assertCountEqual(self, self.g.nodes(), ['a', 'b', 'c'])
six.assertCountEqual(self, self.g.edges(), [('a', 'b'), ('a', 'c')])
self.assertEqual(self.g.parent_node, 'a')
self.assertSetEqual(self.g.children_nodes, {'b', 'c'})
self.assertRaises(ValueError, NaiveBayes, [('a', 'b'), ('b', 'c')])
self.assertRaises(ValueError, NaiveBayes, [('a', 'b'), ('c', 'b')])
self.assertRaises(ValueError, NaiveBayes, [('a', 'b'), ('d', 'e')])
def test_class_init_with_data_nonstring(self):
self.g = NaiveBayes([(1, 2), (1, 3)])
six.assertCountEqual(self, self.g.nodes(), [1, 2, 3])
six.assertCountEqual(self, self.g.edges(), [(1, 2), (1, 3)])
self.assertEqual(self.g.parent_node, 1)
self.assertSetEqual(self.g.children_nodes, {2, 3})
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (2, 3)])
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (3, 2)])
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (3, 4)])
class TestNaiveBayesFit(unittest.TestCase):
def setUp(self):
self.model1 = NaiveBayes()
self.model2 = NaiveBayes([("A", "B")])
def test_fit_model_creation(self):
values = pd.DataFrame(
np.random.randint(low=0, high=2, size=(1000, 5)),
columns=["A", "B", "C", "D", "E"],
)
self.model1.fit(values, "A")
six.assertCountEqual(self, self.model1.nodes(), ["A", "B", "C", "D", "E"])
six.assertCountEqual(
self, self.model1.edges(), [("A", "B"), ("A", "C"), ("A", "D"), ("A", "E")]
)
self.assertEqual(self.model1.parent_node, "A")
self.assertSetEqual(self.model1.children_nodes, {"B", "C", "D", "E"})
self.model2.fit(values)
six.assertCountEqual(self, self.model1.nodes(), ["A", "B", "C", "D", "E"])
six.assertCountEqual(
self, self.model1.edges(), [("A", "B"), ("A", "C"), ("A", "D"), ("A", "E")]
)
self.assertEqual(self.model2.parent_node, "A")
self.assertSetEqual(self.model2.children_nodes, {"B", "C", "D", "E"})
def test_fit_model_creation_exception(self):
values = pd.DataFrame(
np.random.randint(low=0, high=2, size=(1000, 5)),
columns=["A", "B", "C", "D", "E"],
)
values2 = pd.DataFrame(
np.random.randint(low=0, high=2, size=(1000, 3)), columns=["C", "D", "E"]
)
self.assertRaises(ValueError, self.model1.fit, values)
self.assertRaises(ValueError, self.model1.fit, values2)
self.assertRaises(ValueError, self.model2.fit, values2, "A")
def tearDown(self):
del self.model1
del self.model2
class TestNaiveBayesFit(unittest.TestCase):
def setUp(self):
self.model1 = NaiveBayes()
self.model2 = NaiveBayes([('A', 'B')])
def test_fit_model_creation(self):
values = pd.DataFrame(np.random.randint(low=0, high=2, size=(1000, 5)),
columns=['A', 'B', 'C', 'D', 'E'])
self.model1.fit(values, 'A')
six.assertCountEqual(self, self.model1.nodes(),
['A', 'B', 'C', 'D', 'E'])
six.assertCountEqual(self, self.model1.edges(),
[('A', 'B'), ('A', 'C'), ('A', 'D'), ('A', 'E')])
self.assertEqual(self.model1.parent_node, 'A')
self.assertSetEqual(self.model1.children_nodes, {'B', 'C', 'D', 'E'})
self.model2.fit(values)
six.assertCountEqual(self, self.model1.nodes(),
['A', 'B', 'C', 'D', 'E'])
six.assertCountEqual(self, self.model1.edges(),
[('A', 'B'), ('A', 'C'), ('A', 'D'), ('A', 'E')])
self.assertEqual(self.model2.parent_node, 'A')
self.assertSetEqual(self.model2.children_nodes, {'B', 'C', 'D', 'E'})
def test_fit_model_creation_exception(self):
values = pd.DataFrame(np.random.randint(low=0, high=2, size=(1000, 5)),
columns=['A', 'B', 'C', 'D', 'E'])
values2 = pd.DataFrame(np.random.randint(low=0, high=2,
size=(1000, 3)),
columns=['C', 'D', 'E'])
self.assertRaises(ValueError, self.model1.fit, values)
self.assertRaises(ValueError, self.model1.fit, values2)
self.assertRaises(ValueError, self.model2.fit, values2, 'A')
def tearDown(self):
del self.model1
del self.model2
class TestNaiveBayesFit(unittest.TestCase):
def setUp(self):
self.model1 = NaiveBayes()
self.model2 = NaiveBayes([('A','B')])
def test_fit_model_creation(self):
values = pd.DataFrame(np.random.randint(low=0, high=2, size=(1000, 5)),
columns=['A', 'B', 'C', 'D', 'E'])
self.model1.fit(values, 'A')
six.assertCountEqual(self, self.model1.nodes(), ['A', 'B', 'C', 'D', 'E'])
six.assertCountEqual(self, self.model1.edges(), [('A', 'B'), ('A', 'C'), ('A', 'D'),
('A', 'E')])
self.assertEqual(self.model1.parent_node, 'A')
self.assertSetEqual(self.model1.children_nodes, {'B','C','D','E'})
self.model2.fit(values)
six.assertCountEqual(self, self.model1.nodes(), ['A', 'B', 'C', 'D', 'E'])
six.assertCountEqual(self, self.model1.edges(), [('A', 'B'), ('A', 'C'), ('A', 'D'),
('A', 'E')])
self.assertEqual(self.model2.parent_node, 'A')
self.assertSetEqual(self.model2.children_nodes, {'B','C','D','E'})
def test_fit_model_creation_exception(self):
values = pd.DataFrame(np.random.randint(low=0, high=2, size=(1000, 5)),
columns=['A', 'B', 'C', 'D', 'E'])
values2 = pd.DataFrame(np.random.randint(low=0, high=2, size=(1000, 3)),
columns=['C', 'D', 'E'])
self.assertRaises(ValueError, self.model1.fit, values)
self.assertRaises(ValueError, self.model1.fit, values2)
self.assertRaises(ValueError, self.model2.fit, values2, 'A')
def tearDown(self):
del self.model1
del self.model2
def setUp(self):
self.model1 = NaiveBayes()
self.model2 = NaiveBayes([('A','B')])
def setUp(self):
self.G = NaiveBayes()
def test_update_node_parents_bm_constructor(self):
self.g = NaiveBayes([('a', 'b'), ('a', 'c')])
self.assertListEqual(self.g.predecessors('a'), [])
self.assertListEqual(self.g.predecessors('b'), ['a'])
self.assertListEqual(self.g.predecessors('c'), ['a'])
class TestBaseModelCreation(unittest.TestCase):
def setUp(self):
self.G = NaiveBayes()
def test_class_init_without_data(self):
self.assertIsInstance(self.G, nx.DiGraph)
def test_class_init_with_data_string(self):
self.g = NaiveBayes([('a', 'b'), ('a', 'c')])
six.assertCountEqual(self, self.g.nodes(), ['a', 'b', 'c'])
six.assertCountEqual(self, self.g.edges(), [('a', 'b'), ('a', 'c')])
self.assertEqual(self.g.parent_node, 'a')
self.assertSetEqual(self.g.children_nodes, {'b', 'c'})
self.assertRaises(ValueError, NaiveBayes, [('a', 'b'), ('b', 'c')])
self.assertRaises(ValueError, NaiveBayes, [('a', 'b'), ('c', 'b')])
self.assertRaises(ValueError, NaiveBayes, [('a', 'b'), ('d', 'e')])
def test_class_init_with_data_nonstring(self):
self.g = NaiveBayes([(1, 2), (1, 3)])
six.assertCountEqual(self, self.g.nodes(), [1, 2, 3])
six.assertCountEqual(self, self.g.edges(), [(1, 2), (1, 3)])
self.assertEqual(self.g.parent_node, 1)
self.assertSetEqual(self.g.children_nodes, {2, 3})
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (2, 3)])
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (3, 2)])
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (3, 4)])
def test_add_node_string(self):
self.G.add_node('a')
self.assertListEqual(self.G.nodes(), ['a'])
def test_add_node_nonstring(self):
self.G.add_node(1)
self.assertListEqual(self.G.nodes(), [1])
def test_add_nodes_from_string(self):
self.G.add_nodes_from(['a', 'b', 'c', 'd'])
six.assertCountEqual(self, self.G.nodes(), ['a', 'b', 'c', 'd'])
def test_add_nodes_from_non_string(self):
self.G.add_nodes_from([1, 2, 3, 4])
six.assertCountEqual(self, self.G.nodes(), [1, 2, 3, 4])
def test_add_edge_string(self):
self.G.add_edge('a', 'b')
six.assertCountEqual(self, self.G.nodes(), ['a', 'b'])
self.assertListEqual(self.G.edges(), [('a', 'b')])
self.assertEqual(self.G.parent_node, 'a')
self.assertSetEqual(self.G.children_nodes, {'b'})
self.G.add_nodes_from(['c', 'd'])
self.G.add_edge('a', 'c')
self.G.add_edge('a', 'd')
six.assertCountEqual(self, self.G.nodes(), ['a', 'b', 'c', 'd'])
six.assertCountEqual(self, self.G.edges(), [('a', 'b'), ('a', 'c'), ('a', 'd')])
self.assertEqual(self.G.parent_node, 'a')
self.assertSetEqual(self.G.children_nodes, {'b', 'c', 'd'})
self.assertRaises(ValueError, self.G.add_edge, 'b', 'c')
self.assertRaises(ValueError, self.G.add_edge, 'd', 'f')
self.assertRaises(ValueError, self.G.add_edge, 'e', 'f')
self.assertRaises(ValueError, self.G.add_edges_from, [('a', 'e'), ('b', 'f')])
self.assertRaises(ValueError, self.G.add_edges_from, [('b', 'f')])
def test_add_edge_nonstring(self):
self.G.add_edge(1, 2)
six.assertCountEqual(self, self.G.nodes(), [1, 2])
self.assertListEqual(self.G.edges(), [(1, 2)])
self.assertEqual(self.G.parent_node, 1)
self.assertSetEqual(self.G.children_nodes, {2})
self.G.add_nodes_from([3, 4])
self.G.add_edge(1, 3)
self.G.add_edge(1, 4)
six.assertCountEqual(self, self.G.nodes(), [1, 2, 3, 4])
six.assertCountEqual(self, self.G.edges(), [(1, 2), (1, 3), (1, 4)])
self.assertEqual(self.G.parent_node, 1)
self.assertSetEqual(self.G.children_nodes, {2, 3, 4})
self.assertRaises(ValueError, self.G.add_edge, 2, 3)
self.assertRaises(ValueError, self.G.add_edge, 3, 6)
self.assertRaises(ValueError, self.G.add_edge, 5, 6)
self.assertRaises(ValueError, self.G.add_edges_from, [(1, 5), (2, 6)])
self.assertRaises(ValueError, self.G.add_edges_from, [(2, 6)])
def test_add_edge_selfloop(self):
self.assertRaises(ValueError, self.G.add_edge, 'a', 'a')
self.assertRaises(ValueError, self.G.add_edge, 1, 1)
def test_add_edges_from_self_loop(self):
self.assertRaises(ValueError, self.G.add_edges_from,
[('a', 'a')])
def test_update_node_parents_bm_constructor(self):
self.g = NaiveBayes([('a', 'b'), ('a', 'c')])
self.assertListEqual(self.g.predecessors('a'), [])
self.assertListEqual(self.g.predecessors('b'), ['a'])
self.assertListEqual(self.g.predecessors('c'), ['a'])
def test_update_node_parents(self):
self.G.add_nodes_from(['a', 'b', 'c'])
self.G.add_edges_from([('a', 'b'), ('a', 'c')])
self.assertListEqual(self.G.predecessors('a'), [])
self.assertListEqual(self.G.predecessors('b'), ['a'])
self.assertListEqual(self.G.predecessors('c'), ['a'])
def tearDown(self):
del self.G
class PGMNaiveBayes(TextClassifier):
def __add_category(
self, categories: Union[str, List[str], List[(str, str)], Dict[str,
str]]
) -> PGMNaiveBayes:
'''setup the bayes network with a new category entry'''
if type(categories) is str: categories = [categories]
if type(categories) is dict: categories = categories.items()
to_create = False
for category in categories:
if type(category) is str: category = (category, category)
category, index = category
if category not in self.categories:
self.categories[category] = index
self.cardinality = len(self.categories) or 1
to_create = True
if to_create: self.__create_class_cpd()
return self
def __add_token(self, tokens: Union[str, List[str]]) -> PGMNaiveBayes:
if type(tokens) is str: tokens = [tokens]
to_create = []
for token in tokens:
if token not in self.tokens:
to_create.append(token)
self.total_tokens += 1
self.__create_word_cpd(to_create)
return self
def __create_word_cpd(self,
tokens: Union[str, List[str]],
check: bool = True) -> PGMNaiveBayes:
'''Generate the table for the given token node'''
if type(tokens) is str: tokens = [tokens]
cpds = []
for token in tokens:
if token in self.tokens:
self.model.remove_cpds(self.tokens[token])
cpd_word = TabularCPD(
variable=token,
variable_card=2,
evidence=[Data.CATEGORY_NAME],
evidence_card=[self.cardinality],
values=[[0.5 for _ in range(self.cardinality)]] * 2)
self.tokens[token] = cpd_word
cpds.append(cpd_word)
self.model.add_nodes_from(tokens)
self.model.add_edges_from([(Data.CATEGORY_NAME, token)
for token in tokens])
self.model.add_cpds(*cpds)
# if check: self.model.check_model()
return self
def __create_class_cpd(self, check: bool = True) -> PGMNaiveBayes:
'''Generate the table for the category node'''
if self.cpd_class:
self.model.remove_cpds(self.cpd_class)
self.cpd_class = TabularCPD(variable=Data.CATEGORY_NAME,
variable_card=self.cardinality,
values=[[1 / self.cardinality]
for _ in range(self.cardinality)])
self.model.add_cpds(self.cpd_class)
# if check: self.model.check_model()
return self
def __cpd_to_json(self, cpd: TabularCPD) -> Dict:
return {
'variable': cpd.variable,
'variables': cpd.variables,
'variable_card': cpd.variable_card.tolist(),
'values': cpd.values.tolist()
}
def __cpd_from_json(self, cpd: Dict) -> TabularCPD:
return TabularCPD(**cpd)
def reset(self) -> PGMNaiveBayes:
'''Totally reset the Classifier'''
self.categories = {}
self.tokens = {}
self.cardinality = 1
self.total_documents = 0
self.total_tokens = 0
self.cpd_class = None
self.model = NaiveBayes()
self.model.add_node(Data.CATEGORY_NAME)
return self
def token_probability(self, token: str, category: str) -> float:
'''return the probability of a given token to belong a given category'''
probability = self.model.predict_probability(
pd.DataFrame([[1]], columns=[token]))
column = '{}_{}'.format(Data.CATEGORY_NAME,
self.categories.get(category, 0))
return probability[column][0] if column in probability else 0
def category_probability(self, category: str) -> float:
'''return the probability of the given category'''
score = Data.CATEGORY_VALUES.get(category, 0)
elimination = VariableElimination(self.model)
probability = elimination.query(variables=[Data.CATEGORY_NAME])
state = probability.get_state_no(Data.CATEGORY_NAME,
self.categories.get(category, 0))
return probability.values[state]
def word_probability(self, text: str) -> pd.DataFrame:
'''retrive the probability table of the given text without knowing the probability of the category (no evidence): P(C | w1,...,wn)'''
data = Data(text)
elimination = VariableElimination(self.model)
values = [[] for _ in range(self.cardinality)]
for token in data.tokens:
if token not in self.tokens:
for v in values:
v.append(1 / (self.cardinality or 1))
else:
probability = elimination.query(variables=[Data.CATEGORY_NAME],
evidence={
token: 1
}).values
for i in range(len(probability)):
values[i].append(probability[i])
return pd.DataFrame(np.array(values).T,
columns=list(self.categories),
index=data.tokens)
def probability(self, text: str) -> pd.DataFrame:
'''retrive the probability table of the given text knowing the probability of categories: P(C) * P(C | w1,...,wn)'''
data = Data(text)
values = pd.DataFrame(
[[1 if t in data.table else 0 for t in self.tokens]],
columns=self.tokens)
probabilities = self.model.predict_probability(values)
return probabilities.rename(
columns={
'{}_{}'.format(Data.CATEGORY_NAME, v): k
for k, v in self.categories.items()
})
def fit(self,
text: Union[str, Iterable[str], Iterable[Data], pd.DataFrame],
category: Union[str, Iterable[str]] = None) -> TextClassifier:
'''learn probabilities for tokens extracted by the given text'''
data = DataSet.FromAny(text, category)
categories = []
tokens = {}
values = []
for d in data:
categories.append((d.category, d.score))
for token in d.tokens:
tokens[token] = 1
values.append((d.table, d.score))
self.total_documents += 1
tokens = list(tokens)
self.__add_category(categories)
self.__add_token(tokens)
data_values = [[1 if t in v[0] else 0 for t in tokens] + [v[1]]
for v in values]
tokens.append(Data.CATEGORY_NAME)
data_values = pd.DataFrame(data_values, columns=tokens)
self.model.fit(data_values, Data.CATEGORY_NAME)
return self
def words(self, categories: Union[str, Iterable[str]]) -> pd.DataFrame:
'''return a sorted by probability table with tokens as rows and categories as columns, for the given categories'''
elimination = VariableElimination(self.model)
values = [[] for _ in range(self.cardinality)]
for token in self.tokens:
probability = elimination.query(variables=[Data.CATEGORY_NAME],
evidence={
token: 1
}).values
for i in range(len(probability)):
values[i].append(probability[i])
return pd.DataFrame(np.array(values).T,
columns=list(self.categories),
index=list(self.tokens))
def to_json(self) -> Dict:
return {
'categories': self.categories,
'total_documents': self.total_documents,
'tokens': {
c.variable: c.values.tolist()
for c in self.model.get_cpds()
if c.variable != Data.CATEGORY_NAME
},
}
def from_json(self, data: Dict) -> PGMNaiveBayes:
self.total_documents = data.get('total_documents',
self.total_documents)
self.__add_category(data.get('categories', {}))
self.model.remove_cpds(self.cpd_class)
self.cpd_class = TabularCPD(
**data.get('class')) if 'class' in data else self.cpd_class
self.model.add_cpds(self.cpd_class)
tokens = data.get('tokens', {})
self.__add_token(list(tokens))
cpds = {c.variable: c for c in self.model.get_cpds()}
for token, values in tokens.items():
if token in cpds:
cpds[token].values = np.array(values)[0:self.cardinality,
0:self.cardinality]
self.model.check_model()
return self
def __str__(self) -> str:
return 'NaiveBayes<{}, {}>[{}]'.format(self.total_documents,
self.total_tokens,
str.join(', ', self.categories))
def __repr__(self) -> str:
return str(self)
# Print an example of 1 instance in the dataset
print("\nAn example of a person")
print(df.iloc[0])
# Split the data to test and train
test_size = 0.33
print("\nSplitting in to training and test data using: Test size = ", test_size)
data_train, data_test = train_test_split(df, test_size=test_size)
print("training data:", len(data_train))
print("test data:", len(data_test))
#################################################################################
##### Defining the model
#################################################################################
model = NaiveBayes()
# Learning CPDs using Maximum Likelihood Estimators
model.fit(data_train, 'class', estimator=MaximumLikelihoodEstimator)
# Print the CPDs learned
print("\n\n............Overview of our CPDs from the fit...........:")
for cpd in model.get_cpds():
print("CPD of {variable}:".format(variable=cpd.variable))
print(cpd)
print("\n\n............Overview of levels in variables...........:")
for col in df:
print(col,":", len(df[col].unique()) )
#################################################################################
##### Using the model to query
def setUp(self):
self.model1 = NaiveBayes()
self.model2 = NaiveBayes([('A', 'B')])
def setUp(self):
self.model1 = NaiveBayes()
self.model2 = NaiveBayes([("A", "B")])
def setUp(self):
self.G1 = NaiveBayes([('a', 'b'), ('a', 'c'), ('a', 'd'), ('a', 'e')])
self.G2 = NaiveBayes([('d', 'g'), ('d', 'l'), ('d', 's')])
CPD = pickle.load(fp)
with open("RandomColumns.txt", "rb") as fp:
random_columns = pickle.load(fp)
with open("RandomIndices.txt", "rb") as fp:
random_indices = pickle.load(fp)
data = data.iloc[:, random_columns]
column_size = data.shape[1]
#Delete invoices with all zeros from the data
data = data[(data.T != 0).any()]
row_size = data.shape[0]
smallDF = data.iloc[random_indices, :]
smallDF.shape
DictOfModels = {}
for productName in smallDF.columns:
print('Collecting model for {0}'.format(productName))
model = NaiveBayes()
model.add_nodes_from(Nodes[productName])
model.add_edges_from(Edges[productName])
model.add_cpds(*CPD[productName])
DictOfModels[productName] = model
#Save edge ,node, CPD information
PseudoCounts = {}
#Pseudocounts are given (1,1) for uniform
for productName in smallDF.columns:
PseudoCounts[productName] = [1, 1]
except:
print('Existing model not found')
#Select random invoice (2000) and products (50)
seed(0)
column_size = data.shape[1]
random_columns = sample(range(column_size), 100)
class TestNaiveBayesMethods(unittest.TestCase):
def setUp(self):
self.G1 = NaiveBayes([('a', 'b'), ('a', 'c'), ('a', 'd'), ('a', 'e')])
self.G2 = NaiveBayes([('d', 'g'), ('d', 'l'), ('d', 's')])
def test_local_independencies(self):
self.assertListEqual(self.G1.local_independencies('a'), [None])
self.assertListEqual(self.G1.local_independencies('b'),
[Independencies(['b', ['e', 'c', 'd'], 'a'])])
self.assertListEqual(self.G1.local_independencies('c'),
[Independencies(['c', ['e', 'b', 'd'], 'a'])])
self.assertListEqual(self.G1.local_independencies('d'),
[Independencies(['d', ['b', 'c', 'e'], 'a'])])
def test_active_trail_nodes(self):
self.assertListEqual(sorted(self.G2.active_trail_nodes('d')),
['d', 'g', 'l', 's'])
self.assertListEqual(sorted(self.G2.active_trail_nodes('g')),
['d', 'g', 'l', 's'])
self.assertListEqual(sorted(self.G2.active_trail_nodes('l')),
['d', 'g', 'l', 's'])
self.assertListEqual(sorted(self.G2.active_trail_nodes('s')),
['d', 'g', 'l', 's'])
def test_active_trail_nodes_args(self):
self.assertListEqual(
sorted(self.G2.active_trail_nodes('d', observed='g')),
['d', 'l', 's'])
self.assertListEqual(
sorted(self.G2.active_trail_nodes('l', observed='g')),
['d', 'l', 's'])
self.assertListEqual(
sorted(self.G2.active_trail_nodes('s', observed=['g', 'l'])),
['d', 's'])
self.assertListEqual(
sorted(self.G2.active_trail_nodes('s', observed=['d', 'l'])),
['s'])
def tearDown(self):
del self.G1
del self.G2
def setUp(self):
self.G = NaiveBayes()
class TestBaseModelCreation(unittest.TestCase):
def setUp(self):
self.G = NaiveBayes()
def test_class_init_without_data(self):
self.assertIsInstance(self.G, nx.DiGraph)
def test_class_init_with_data_string(self):
self.g = NaiveBayes([('a', 'b'), ('a', 'c')])
six.assertCountEqual(self, self.g.nodes(), ['a', 'b', 'c'])
six.assertCountEqual(self, self.g.edges(), [('a', 'b'), ('a', 'c')])
self.assertEqual(self.g.parent_node, 'a')
self.assertSetEqual(self.g.children_nodes, {'b', 'c'})
self.assertRaises(ValueError, NaiveBayes, [('a', 'b'), ('b', 'c')])
self.assertRaises(ValueError, NaiveBayes, [('a', 'b'), ('c', 'b')])
self.assertRaises(ValueError, NaiveBayes, [('a', 'b'), ('d', 'e')])
def test_class_init_with_data_nonstring(self):
self.g = NaiveBayes([(1, 2), (1, 3)])
six.assertCountEqual(self, self.g.nodes(), [1, 2, 3])
six.assertCountEqual(self, self.g.edges(), [(1, 2), (1, 3)])
self.assertEqual(self.g.parent_node, 1)
self.assertSetEqual(self.g.children_nodes, {2, 3})
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (2, 3)])
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (3, 2)])
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (3, 4)])
def test_add_node_string(self):
self.G.add_node('a')
self.assertListEqual(self.G.nodes(), ['a'])
def test_add_node_nonstring(self):
self.G.add_node(1)
self.assertListEqual(self.G.nodes(), [1])
def test_add_nodes_from_string(self):
self.G.add_nodes_from(['a', 'b', 'c', 'd'])
six.assertCountEqual(self, self.G.nodes(), ['a', 'b', 'c', 'd'])
def test_add_nodes_from_non_string(self):
self.G.add_nodes_from([1, 2, 3, 4])
six.assertCountEqual(self, self.G.nodes(), [1, 2, 3, 4])
def test_add_edge_string(self):
self.G.add_edge('a', 'b')
six.assertCountEqual(self, self.G.nodes(), ['a', 'b'])
self.assertListEqual(self.G.edges(), [('a', 'b')])
self.assertEqual(self.G.parent_node, 'a')
self.assertSetEqual(self.G.children_nodes, {'b'})
self.G.add_nodes_from(['c', 'd'])
self.G.add_edge('a', 'c')
self.G.add_edge('a', 'd')
six.assertCountEqual(self, self.G.nodes(), ['a', 'b', 'c', 'd'])
six.assertCountEqual(self, self.G.edges(), [('a', 'b'), ('a', 'c'),
('a', 'd')])
self.assertEqual(self.G.parent_node, 'a')
self.assertSetEqual(self.G.children_nodes, {'b', 'c', 'd'})
self.assertRaises(ValueError, self.G.add_edge, 'b', 'c')
self.assertRaises(ValueError, self.G.add_edge, 'd', 'f')
self.assertRaises(ValueError, self.G.add_edge, 'e', 'f')
self.assertRaises(ValueError, self.G.add_edges_from, [('a', 'e'),
('b', 'f')])
self.assertRaises(ValueError, self.G.add_edges_from, [('b', 'f')])
def test_add_edge_nonstring(self):
self.G.add_edge(1, 2)
six.assertCountEqual(self, self.G.nodes(), [1, 2])
self.assertListEqual(self.G.edges(), [(1, 2)])
self.assertEqual(self.G.parent_node, 1)
self.assertSetEqual(self.G.children_nodes, {2})
self.G.add_nodes_from([3, 4])
self.G.add_edge(1, 3)
self.G.add_edge(1, 4)
six.assertCountEqual(self, self.G.nodes(), [1, 2, 3, 4])
six.assertCountEqual(self, self.G.edges(), [(1, 2), (1, 3), (1, 4)])
self.assertEqual(self.G.parent_node, 1)
self.assertSetEqual(self.G.children_nodes, {2, 3, 4})
self.assertRaises(ValueError, self.G.add_edge, 2, 3)
self.assertRaises(ValueError, self.G.add_edge, 3, 6)
self.assertRaises(ValueError, self.G.add_edge, 5, 6)
self.assertRaises(ValueError, self.G.add_edges_from, [(1, 5), (2, 6)])
self.assertRaises(ValueError, self.G.add_edges_from, [(2, 6)])
def test_add_edge_selfloop(self):
self.assertRaises(ValueError, self.G.add_edge, 'a', 'a')
self.assertRaises(ValueError, self.G.add_edge, 1, 1)
def test_add_edges_from_self_loop(self):
self.assertRaises(ValueError, self.G.add_edges_from, [('a', 'a')])
def test_update_node_parents_bm_constructor(self):
self.g = NaiveBayes([('a', 'b'), ('a', 'c')])
self.assertListEqual(self.g.predecessors('a'), [])
self.assertListEqual(self.g.predecessors('b'), ['a'])
self.assertListEqual(self.g.predecessors('c'), ['a'])
def test_update_node_parents(self):
self.G.add_nodes_from(['a', 'b', 'c'])
self.G.add_edges_from([('a', 'b'), ('a', 'c')])
self.assertListEqual(self.G.predecessors('a'), [])
self.assertListEqual(self.G.predecessors('b'), ['a'])
self.assertListEqual(self.G.predecessors('c'), ['a'])
def tearDown(self):
del self.G
class TestNaiveBayesMethods(unittest.TestCase):
def setUp(self):
self.G1 = NaiveBayes([("a", "b"), ("a", "c"), ("a", "d"), ("a", "e")])
self.G2 = NaiveBayes([("d", "g"), ("d", "l"), ("d", "s")])
def test_local_independencies(self):
self.assertEqual(self.G1.local_independencies("a"), Independencies())
self.assertEqual(
self.G1.local_independencies("b"),
Independencies(["b", ["e", "c", "d"], "a"]),
)
self.assertEqual(
self.G1.local_independencies("c"),
Independencies(["c", ["e", "b", "d"], "a"]),
)
self.assertEqual(
self.G1.local_independencies("d"),
Independencies(["d", ["b", "c", "e"], "a"]),
)
def test_active_trail_nodes(self):
self.assertListEqual(
sorted(self.G2.active_trail_nodes("d")), ["d", "g", "l", "s"]
)
self.assertListEqual(
sorted(self.G2.active_trail_nodes("g")), ["d", "g", "l", "s"]
)
self.assertListEqual(
sorted(self.G2.active_trail_nodes("l")), ["d", "g", "l", "s"]
)
self.assertListEqual(
sorted(self.G2.active_trail_nodes("s")), ["d", "g", "l", "s"]
)
def test_active_trail_nodes_args(self):
self.assertListEqual(
sorted(self.G2.active_trail_nodes("d", observed="g")), ["d", "l", "s"]
)
self.assertListEqual(
sorted(self.G2.active_trail_nodes("l", observed="g")), ["d", "l", "s"]
)
self.assertListEqual(
sorted(self.G2.active_trail_nodes("s", observed=["g", "l"])), ["d", "s"]
)
self.assertListEqual(
sorted(self.G2.active_trail_nodes("s", observed=["d", "l"])), ["s"]
)
def test_get_ancestors_of(self):
self.assertListEqual(sorted(self.G1._get_ancestors_of("b")), ["a", "b"])
self.assertListEqual(sorted(self.G1._get_ancestors_of("e")), ["a", "e"])
self.assertListEqual(sorted(self.G1._get_ancestors_of("a")), ["a"])
self.assertListEqual(
sorted(self.G1._get_ancestors_of(["b", "e"])), ["a", "b", "e"]
)
def tearDown(self):
del self.G1
del self.G2
def setUp(self):
self.G1 = NaiveBayes([("a", "b"), ("a", "c"), ("a", "d"), ("a", "e")])
self.G2 = NaiveBayes([("d", "g"), ("d", "l"), ("d", "s")])
class TestBaseModelCreation(unittest.TestCase):
def setUp(self):
self.G = NaiveBayes()
def test_class_init_without_data(self):
self.assertIsInstance(self.G, nx.DiGraph)
def test_class_init_with_data_string(self):
self.g = NaiveBayes([("a", "b"), ("a", "c")])
six.assertCountEqual(self, list(self.g.nodes()), ["a", "b", "c"])
six.assertCountEqual(self, list(self.g.edges()), [("a", "b"), ("a", "c")])
self.assertEqual(self.g.parent_node, "a")
self.assertSetEqual(self.g.children_nodes, {"b", "c"})
self.assertRaises(ValueError, NaiveBayes, [("a", "b"), ("b", "c")])
self.assertRaises(ValueError, NaiveBayes, [("a", "b"), ("c", "b")])
self.assertRaises(ValueError, NaiveBayes, [("a", "b"), ("d", "e")])
def test_class_init_with_data_nonstring(self):
self.g = NaiveBayes([(1, 2), (1, 3)])
six.assertCountEqual(self, list(self.g.nodes()), [1, 2, 3])
six.assertCountEqual(self, list(self.g.edges()), [(1, 2), (1, 3)])
self.assertEqual(self.g.parent_node, 1)
self.assertSetEqual(self.g.children_nodes, {2, 3})
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (2, 3)])
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (3, 2)])
self.assertRaises(ValueError, NaiveBayes, [(1, 2), (3, 4)])
def test_add_node_string(self):
self.G.add_node("a")
self.assertListEqual(list(self.G.nodes()), ["a"])
def test_add_node_nonstring(self):
self.G.add_node(1)
self.assertListEqual(list(self.G.nodes()), [1])
def test_add_nodes_from_string(self):
self.G.add_nodes_from(["a", "b", "c", "d"])
six.assertCountEqual(self, list(self.G.nodes()), ["a", "b", "c", "d"])
def test_add_nodes_from_non_string(self):
self.G.add_nodes_from([1, 2, 3, 4])
six.assertCountEqual(self, list(self.G.nodes()), [1, 2, 3, 4])
def test_add_edge_string(self):
self.G.add_edge("a", "b")
six.assertCountEqual(self, list(self.G.nodes()), ["a", "b"])
self.assertListEqual(list(self.G.edges()), [("a", "b")])
self.assertEqual(self.G.parent_node, "a")
self.assertSetEqual(self.G.children_nodes, {"b"})
self.G.add_nodes_from(["c", "d"])
self.G.add_edge("a", "c")
self.G.add_edge("a", "d")
six.assertCountEqual(self, list(self.G.nodes()), ["a", "b", "c", "d"])
six.assertCountEqual(
self, list(self.G.edges()), [("a", "b"), ("a", "c"), ("a", "d")]
)
self.assertEqual(self.G.parent_node, "a")
self.assertSetEqual(self.G.children_nodes, {"b", "c", "d"})
self.assertRaises(ValueError, self.G.add_edge, "b", "c")
self.assertRaises(ValueError, self.G.add_edge, "d", "f")
self.assertRaises(ValueError, self.G.add_edge, "e", "f")
self.assertRaises(ValueError, self.G.add_edges_from, [("a", "e"), ("b", "f")])
self.assertRaises(ValueError, self.G.add_edges_from, [("b", "f")])
def test_add_edge_nonstring(self):
self.G.add_edge(1, 2)
six.assertCountEqual(self, list(self.G.nodes()), [1, 2])
self.assertListEqual(list(self.G.edges()), [(1, 2)])
self.assertEqual(self.G.parent_node, 1)
self.assertSetEqual(self.G.children_nodes, {2})
self.G.add_nodes_from([3, 4])
self.G.add_edge(1, 3)
self.G.add_edge(1, 4)
six.assertCountEqual(self, list(self.G.nodes()), [1, 2, 3, 4])
six.assertCountEqual(self, list(self.G.edges()), [(1, 2), (1, 3), (1, 4)])
self.assertEqual(self.G.parent_node, 1)
self.assertSetEqual(self.G.children_nodes, {2, 3, 4})
self.assertRaises(ValueError, self.G.add_edge, 2, 3)
self.assertRaises(ValueError, self.G.add_edge, 3, 6)
self.assertRaises(ValueError, self.G.add_edge, 5, 6)
self.assertRaises(ValueError, self.G.add_edges_from, [(1, 5), (2, 6)])
self.assertRaises(ValueError, self.G.add_edges_from, [(2, 6)])
def test_add_edge_selfloop(self):
self.assertRaises(ValueError, self.G.add_edge, "a", "a")
self.assertRaises(ValueError, self.G.add_edge, 1, 1)
def test_add_edges_from_self_loop(self):
self.assertRaises(ValueError, self.G.add_edges_from, [("a", "a")])
def test_update_node_parents_bm_constructor(self):
self.g = NaiveBayes([("a", "b"), ("a", "c")])
self.assertListEqual(list(self.g.predecessors("a")), [])
self.assertListEqual(list(self.g.predecessors("b")), ["a"])
self.assertListEqual(list(self.g.predecessors("c")), ["a"])
def test_update_node_parents(self):
self.G.add_nodes_from(["a", "b", "c"])
self.G.add_edges_from([("a", "b"), ("a", "c")])
self.assertListEqual(list(self.G.predecessors("a")), [])
self.assertListEqual(list(self.G.predecessors("b")), ["a"])
self.assertListEqual(list(self.G.predecessors("c")), ["a"])
def tearDown(self):
del self.G
def test_update_node_parents_bm_constructor(self):
self.g = NaiveBayes([('a', 'b'), ('a', 'c')])
self.assertListEqual(self.g.predecessors('a'), [])
self.assertListEqual(self.g.predecessors('b'), ['a'])
self.assertListEqual(self.g.predecessors('c'), ['a'])
def test_update_node_parents_bm_constructor(self):
self.g = NaiveBayes([("a", "b"), ("a", "c")])
self.assertListEqual(list(self.g.predecessors("a")), [])
self.assertListEqual(list(self.g.predecessors("b")), ["a"])
self.assertListEqual(list(self.g.predecessors("c")), ["a"])
def setUp(self):
self.G1 = NaiveBayes([('a', 'b'), ('a', 'c'),
('a', 'd'), ('a', 'e')])
self.G2 = NaiveBayes([('d', 'g'), ('d', 'l'),
('d', 's')])
# Split the data to test and train
print("Data set size:", len(df))
print("\n\n............Splitting the data in test and train...........:\n")
test_size = 0.33
print("Test size = ", test_size)
data_train, data_test = train_test_split(df, test_size=test_size)
print("training data:", len(data_train))
print("test data:", len(data_test))
input("\n ")
#################################################################################
##### Defining the model
#################################################################################
model = NaiveBayes()
# Learning CPDs using Maximum Likelihood Estimators
model.fit(data_train, 'class', estimator=MaximumLikelihoodEstimator)
# Print the CPDs learned
print("\n\n............Selected CPDs from the fit...........:\n")
print('CPD: class (parent of all nodes)')
print(model.get_cpds('class'))
input("\n ")
print('\nCPD: sex')
print(model.get_cpds('sex'))
# print(model.get_cpds('race'))
input("\n ")
# print("\n\n............Overview of levels in variables...........:\n")
from sklearn.metrics import precision_score
from sklearn.metrics import f1_score
from sklearn.preprocessing import LabelEncoder
col_names = pd.read_csv('data/names.csv') # 'data/names.csv'
data = pd.read_csv('data/breast-cancer-wisconsin.data', names=col_names.columns)
data = data[data["bare_nuclei"] != '?']
data.set_index('id', inplace=True) #stop the model from using id as a node
train, test = train_test_split(data, test_size=0.2, random_state=0)
Y_test = test['class']
test = test.drop(['class'], axis=1)
#fit model
model = NaiveBayes()
model.fit(train, 'class')
print("Naive Bayes edges: ", model.edges())
#make predictions
Y_pred = model.predict(test)
#Convert Labels so we can use sklearn function to evaluate our model
labelencoder = LabelEncoder()
Y_test = labelencoder.fit_transform(Y_test.values.ravel())
Y_pred = labelencoder.fit_transform(Y_pred.values.ravel())
# Output results
accuracy = accuracy_score(Y_test, Y_pred)
precision = precision_score(Y_test, Y_pred)
f1 = f1_score(Y_test, Y_pred)
