def estimate(self):
"""
Estimates the `BayesianModel` structure that fits best to the given data set,
according to the scoring method supplied in the constructor.
Exhaustively searches through all models. Only estimates network structure, no parametrization.
Returns
-------
model: `BayesianModel` instance
A `BayesianModel` with maximal score.
Examples
--------
>>> import pandas as pd
>>> import numpy as np
>>> from pgmpy.estimators import ExhaustiveSearch
>>> # create random data sample with 3 variables, where B and C are identical:
>>> data = pd.DataFrame(np.random.randint(0, 5, size=(5000, 2)), columns=list('AB'))
>>> data['C'] = data['B']
>>> est = ExhaustiveSearch(data)
>>> best_model = est.estimate()
>>> best_model
<pgmpy.models.BayesianModel.BayesianModel object at 0x7f695c535470>
>>> best_model.edges()
[('B', 'C')]
"""
best_dag = max(self.all_dags(), key=self.scoring_method.score)
best_model = BayesianModel()
best_model.add_nodes_from(sorted(best_dag.nodes()))
best_model.add_edges_from(sorted(best_dag.edges()))
return best_model
def test_score_titanic(self):
scorer = BicScore(self.titanic_data2)
titanic = BayesianModel([("Sex", "Survived"), ("Pclass", "Survived")])
self.assertAlmostEqual(scorer.score(titanic), -1896.7250012840179)
titanic2 = BayesianModel([("Pclass", "Sex")])
titanic2.add_nodes_from(["Sex", "Survived", "Pclass"])
self.assertLess(scorer.score(titanic2), scorer.score(titanic))
def Hill_Climb_Search(data,state_names): epsilon = 1e-14 nodes = state_names.keys() start = BayesianModel() start.add_nodes_from(nodes) current_model = start while True: best_score_delta = 0 best_operation = None for operation, score_delta in operations(data,current_model, state_names): if score_delta > best_score_delta: best_operation = operation best_score_delta = score_delta if best_operation is None or best_score_delta < epsilon: break elif best_operation[0] == 'add': current_model.add_edge(*best_operation[1]) elif best_operation[0] == 'delete': current_model.remove_edge(*best_operation[1]) elif best_operation[0] == 'reverse': X, Y = best_operation[1] current_model.remove_edge(X, Y) current_model.add_edge(Y, X) print 'Iteration:' print current_model.edges() print 'Score: ',best_score_delta print 'Best operation: ',best_operation print current_model.edges() return current_model
def get_model(self):
"""
Returns an instance of Bayesian Model.
"""
model = BayesianModel()
model.add_nodes_from(self.variables)
model.add_edges_from(self.edges)
model.name = self.model_name
tabular_cpds = []
for var, values in self.variable_CPD.items():
evidence = values['CONDSET'] if 'CONDSET' in values else []
cpd = values['DPIS']
evidence_card = values[
'CARDINALITY'] if 'CARDINALITY' in values else []
states = self.variables[var]['STATES']
cpd = TabularCPD(var,
len(states),
cpd,
evidence=evidence,
evidence_card=evidence_card)
tabular_cpds.append(cpd)
model.add_cpds(*tabular_cpds)
for var, properties in self.variables.items():
model.node[var] = properties
return model
def estimate(self):
"""
Estimates the `BayesianModel` structure that fits best to the given data set,
according to the scoring method supplied in the constructor.
Exhaustively searches through all models. Only estimates network structure, no parametrization.
Returns
-------
model: `BayesianModel` instance
A `BayesianModel` with maximal score.
Examples
--------
>>> import pandas as pd
>>> import numpy as np
>>> from pgmpy.estimators import ExhaustiveSearch
>>> # create random data sample with 3 variables, where B and C are identical:
>>> data = pd.DataFrame(np.random.randint(0, 5, size=(5000, 2)), columns=list('AB'))
>>> data['C'] = data['B']
>>> est = ExhaustiveSearch(data)
>>> best_model = est.estimate()
>>> best_model
<pgmpy.models.BayesianModel.BayesianModel object at 0x7f695c535470>
>>> best_model.edges()
[('B', 'C')]
"""
best_dag = max(self.all_dags(), key=self.scoring_method.score)
best_model = BayesianModel()
best_model.add_nodes_from(sorted(best_dag.nodes()))
best_model.add_edges_from(sorted(best_dag.edges()))
return best_model
def _train_bn(self):
model = BayesianModel(self._dag.edges)
model.add_nodes_from(self._dag.nodes)
model.fit(self._get_training_data(),
BayesianEstimator,
prior_type='BDeu')
return model
def get_model(self):
"""
Returns an instance of Bayesian Model.
"""
model = BayesianModel()
model.add_nodes_from(self.variables)
model.add_edges_from(self.edges)
model.name = self.model_name
tabular_cpds = []
for var, values in self.variable_CPD.items():
evidence = values["CONDSET"] if "CONDSET" in values else []
cpd = values["DPIS"]
evidence_card = values[
"CARDINALITY"] if "CARDINALITY" in values else []
states = self.variables[var]["STATES"]
cpd = TabularCPD(var,
len(states),
cpd,
evidence=evidence,
evidence_card=evidence_card)
tabular_cpds.append(cpd)
model.add_cpds(*tabular_cpds)
if nx.__version__.startswith("1"):
for var, properties in self.variables.items():
model.nodes[var] = properties
else:
for var, properties in self.variables.items():
model._node[var] = properties
return model
def get_model(self):
"""
Returns the fitted bayesian model
Example
----------
>>> from pgmpy.readwrite import BIFReader
>>> reader = BIFReader("bif_test.bif")
>>> reader.get_model()
<pgmpy.models.BayesianModel.BayesianModel object at 0x7f20af154320>
"""
try:
model = BayesianModel(self.variable_edges)
model.name = self.network_name
model.add_nodes_from(self.variable_names)
tabular_cpds = []
for var in sorted(self.variable_cpds.keys()):
values = self.variable_cpds[var]
cpd = TabularCPD(var, len(self.variable_states[var]), values,
evidence=self.variable_parents[var],
evidence_card=[len(self.variable_states[evidence_var])
for evidence_var in self.variable_parents[var]])
tabular_cpds.append(cpd)
model.add_cpds(*tabular_cpds)
for node, properties in self.variable_properties.items():
for prop in properties:
prop_name, prop_value = map(lambda t: t.strip(), prop.split('='))
model.node[node][prop_name] = prop_value
return model
except AttributeError:
raise AttributeError('First get states of variables, edges, parents and network name')
def test_score_titanic(self):
scorer = K2Score(self.titanic_data2)
titanic = BayesianModel([("Sex", "Survived"), ("Pclass", "Survived")])
self.assertAlmostEqual(scorer.score(titanic), -1891.0630673606006)
titanic2 = BayesianModel([("Pclass", "Sex"), ])
titanic2.add_nodes_from(["Sex", "Survived", "Pclass"])
self.assertLess(scorer.score(titanic2), scorer.score(titanic))
def test_score_titanic(self):
scorer = BdeuScore(self.titanic_data2, equivalent_sample_size=25)
titanic = BayesianModel([("Sex", "Survived"), ("Pclass", "Survived")])
self.assertAlmostEqual(scorer.score(titanic), -1892.7383393910427)
titanic2 = BayesianModel([("Pclass", "Sex")])
titanic2.add_nodes_from(["Sex", "Survived", "Pclass"])
self.assertLess(scorer.score(titanic2), scorer.score(titanic))
def get_model(self):
model = BayesianModel()
model.add_nodes_from(self.variables)
model.add_edges_from(self.edge_list)
model.name = self.network_name
tabular_cpds = []
for var, values in self.variable_CPD.items():
evidence_card = [
len(self.variable_states[evidence_var])
for evidence_var in self.variable_parents[var]
]
cpd = TabularCPD(
var,
len(self.variable_states[var]),
values,
evidence=self.variable_parents[var],
evidence_card=evidence_card,
state_names=self.get_states(),
)
tabular_cpds.append(cpd)
model.add_cpds(*tabular_cpds)
for node, properties in self.variable_property.items():
for prop in properties:
if prop is not None:
prop_name, prop_value = map(lambda t: t.strip(), prop.split("="))
model.nodes[node][prop_name] = prop_value
return model
def single_bayes_net(df, independent_vars, dependent_vars):
model = BayesianModel()
model.add_nodes_from(independent_vars)
for independent_var in independent_vars:
for dependent_var in dependent_vars:
model.add_edge(independent_var, dependent_var)
model.fit(df)
return model
def get_model(self):
"""
Returns the model instance of the ProbModel.
Return
---------------
model: an instance of BayesianModel.
Examples
-------
>>> reader = ProbModelXMLReader()
>>> reader.get_model()
"""
if self.probnet.get("type") == "BayesianNetwork":
model = BayesianModel()
model.add_nodes_from(self.probnet["Variables"].keys())
model.add_edges_from(self.probnet["edges"].keys())
tabular_cpds = []
cpds = self.probnet["Potentials"]
for cpd in cpds:
var = list(cpd["Variables"].keys())[0]
states = self.probnet["Variables"][var]["States"]
evidence = cpd["Variables"][var]
evidence_card = [
len(self.probnet["Variables"][evidence_var]["States"])
for evidence_var in evidence
]
arr = list(map(float, cpd["Values"].split()))
values = np.array(arr)
values = values.reshape((len(states), values.size // len(states)))
tabular_cpds.append(
TabularCPD(var, len(states), values, evidence, evidence_card)
)
model.add_cpds(*tabular_cpds)
variables = model.nodes()
for var in variables:
for prop_name, prop_value in self.probnet["Variables"][var].items():
model.nodes[var][prop_name] = prop_value
edges = model.edges()
if nx.__version__.startswith("1"):
for edge in edges:
for prop_name, prop_value in self.probnet["edges"][edge].items():
model.edge[edge[0]][edge[1]][prop_name] = prop_value
else:
for edge in edges:
for prop_name, prop_value in self.probnet["edges"][edge].items():
model.adj[edge[0]][edge[1]][prop_name] = prop_value
return model
else:
raise ValueError("Please specify only Bayesian Network.")
def create_bayes_net(file, keep_atts, edges):
atts = pd.read_csv(file)
atts = atts[keep_atts]
graph = BayesianModel()
graph.add_nodes_from(atts.columns)
# defining the structure of edges
graph.add_edges_from(edges)
# fit estimates the CPD tables for the given structure
graph.fit(atts)
return graph
class TestBayesianModelFitPredict(unittest.TestCase):
def setUp(self):
self.model_disconnected = BayesianModel()
self.model_disconnected.add_nodes_from(['A', 'B', 'C', 'D', 'E'])
self.model_connected = BayesianModel([('A', 'B'), ('C', 'B'),
('C', 'D'), ('B', 'E')])
def test_disconnected_fit(self):
values = pd.DataFrame(np.random.randint(low=0, high=2, size=(1000, 5)),
columns=['A', 'B', 'C', 'D', 'E'])
self.model_disconnected.fit(values)
for node in ['A', 'B', 'C', 'D', 'E']:
cpd = self.model_disconnected.get_cpds(node)
self.assertEqual(cpd.variable, node)
np_test.assert_array_equal(cpd.cardinality, np.array([2]))
value = (values.ix[:, node].value_counts() /
values.ix[:, node].value_counts().sum())
value = value.reindex(sorted(value.index)).values
np_test.assert_array_equal(cpd.values, value)
def test_connected_predict(self):
np.random.seed(42)
values = pd.DataFrame(np.random.randint(low=0, high=2, size=(1000, 5)),
columns=['A', 'B', 'C', 'D', 'E'])
fit_data = values[:800]
predict_data = values[800:].copy()
self.model_connected.fit(fit_data)
self.assertRaises(ValueError, self.model_connected.predict,
predict_data)
predict_data.drop('E', axis=1, inplace=True)
e_predict = self.model_connected.predict(predict_data)
np_test.assert_array_equal(
e_predict.values.ravel(),
np.array([
1, 0, 1, 0, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 0, 0, 1, 1, 1, 0,
0, 0, 1, 1, 0, 1, 0, 0, 1, 0, 0, 0, 0, 1, 1, 1, 0, 1, 1, 0, 0,
0, 0, 0, 1, 1, 0, 1, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 1,
1, 1, 1, 1, 0, 0, 1, 1, 0, 1, 1, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1,
1, 1, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 1, 1, 1, 0, 0, 1,
1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 1, 1, 0, 1, 1, 0, 0, 1, 1,
1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 1, 1,
1, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 1, 0, 1, 0,
1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0
]))
def tearDown(self):
del self.model_connected
del self.model_disconnected
class TestBayesianModelFitPredict(unittest.TestCase):
def setUp(self):
self.model_disconnected = BayesianModel()
self.model_disconnected.add_nodes_from(['A', 'B', 'C', 'D', 'E'])
self.model_connected = BayesianModel([('A', 'B'), ('C', 'B'), ('C', 'D'), ('B', 'E')])
def test_disconnected_fit(self):
values = pd.DataFrame(np.random.randint(low=0, high=2, size=(1000, 5)),
columns=['A', 'B', 'C', 'D', 'E'])
self.model_disconnected.fit(values)
for node in ['A', 'B', 'C', 'D', 'E']:
cpd = self.model_disconnected.get_cpds(node)
self.assertEqual(cpd.variable, node)
np_test.assert_array_equal(cpd.cardinality, np.array([2]))
value = (values.ix[:, node].value_counts() /
values.ix[:, node].value_counts().sum())
value = value.reindex(sorted(value.index)).values
np_test.assert_array_equal(cpd.values, value)
def test_connected_predict(self):
np.random.seed(42)
values = pd.DataFrame(np.random.randint(low=0, high=2, size=(1000, 5)),
columns=['A', 'B', 'C', 'D', 'E'])
fit_data = values[:800]
predict_data = values[800:].copy()
self.model_connected.fit(fit_data)
self.assertRaises(ValueError, self.model_connected.predict, predict_data)
predict_data.drop('E', axis=1, inplace=True)
e_predict = self.model_connected.predict(predict_data)
np_test.assert_array_equal(e_predict.values.ravel(),
np.array([1, 0, 1, 0, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1,
1, 0, 0, 1, 1, 1, 0, 0, 0, 1, 1, 0, 1, 0,
0, 1, 0, 0, 0, 0, 1, 1, 1, 0, 1, 1, 0, 0,
0, 0, 0, 1, 1, 0, 1, 1, 0, 0, 0, 1, 0, 1,
0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 1,
1, 0, 1, 1, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1,
1, 1, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 0, 0,
1, 1, 1, 1, 0, 0, 1, 1, 0, 0, 0, 0, 1, 1,
0, 0, 0, 1, 0, 1, 1, 0, 1, 1, 0, 0, 1, 1,
1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1,
1, 0, 1, 0, 1, 1, 1, 1, 0, 0, 1, 0, 1, 1,
0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 1, 0, 1, 0,
1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 0, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 0, 1,
1, 1, 1, 0]))
def tearDown(self):
del self.model_connected
del self.model_disconnected
def create_model_and_inference():
dep_df = pd.read_csv('dependencies.csv', sep=';')
def connect(df, source, edgelist):
source_df = df[df['Column2'] == source]
for col in source_df.iloc[0, 3:len(source_df.columns)]:
target_df = df[df['Column1'] == col]['Column2']
if not target_df.empty:
target = target_df.item()
if not (target, source) in edgelist:
edgelist.append((source, target))
connect(df, target, edgelist)
edges = []
connect(dep_df, 'myproximus-usage', edges)
edges = [(t[1], t[0]) for t in edges]
nodes = set(itertools.chain.from_iterable(edges))
nodes_df = dep_df.iloc[:, 1].to_frame()
nodes_df = nodes_df[nodes_df['Column2'].isin(nodes)]
nodes_df['0'] = pd.DataFrame(data=np.random.randint(0, 2, size=64).T)
nodes_df['1'] = pd.DataFrame(data=np.random.randint(0, 2, size=64).T)
nodes_df['2'] = pd.DataFrame(data=np.random.randint(0, 2, size=64).T)
nodes_df['3'] = pd.DataFrame(data=np.random.randint(0, 2, size=64).T)
nodes_df['4'] = pd.DataFrame(data=np.random.randint(0, 2, size=64).T)
nodes_df['5'] = pd.DataFrame(data=np.random.randint(0, 2, size=64).T)
nodes_df['6'] = pd.DataFrame(data=np.random.randint(0, 2, size=64).T)
nodes_df['7'] = pd.DataFrame(data=np.random.randint(0, 2, size=64).T)
nodes_df['8'] = pd.DataFrame(data=np.random.randint(0, 2, size=64).T)
nodes_df['9'] = pd.DataFrame(data=np.random.randint(0, 2, size=64).T)
nodes_df['10'] = pd.DataFrame(data=np.random.randint(0, 2, size=64).T)
nodes_df = nodes_df.set_index('Column2').transpose()
model = BayesianModel()
model.add_nodes_from(nodes)
for edge in edges:
try:
model.add_edge(edge[0], edge[1])
except:
print('WARNING: tried to add edge which forms loop: ' + str(edge))
model.fit(nodes_df, estimator=BayesianEstimator, prior_type="BDeu")
# for cpd in model.get_cpds():
# print(cpd)
draw_network(model.nodes(), model.edges(), {}, [])
return model, VariableElimination(model)
def bayes_net_from_populational_data(data, independent_vars,
dependent_vars):
model = BayesianModel()
model.add_nodes_from(independent_vars)
for independent_var in independent_vars:
for dependent_var in dependent_vars:
model.add_edge(independent_var, dependent_var)
cpd_list = []
state_names = BayesNetHelper.get_state_names_from_df(
data, independent_vars | dependent_vars)
for node in independent_vars | dependent_vars:
cpd = BayesNetHelper.compute_cpd(model, node, data, state_names)
cpd_list.append(cpd)
model.add_cpds(*cpd_list)
return model
def fully_connected_model(nodes=None):
if not nodes:
nodes = [BOREDOM, DESIRE, MOBILE, MOTOR_HYPO, LEFT_ARM]
network = BayesianModel()
network.add_nodes_from(nodes)
for hypo in nodes:
if 'hypo' in hypo:
for obs in nodes:
if 'obs' in obs or 'motor' in obs:
network.add_edge(u=hypo, v=obs)
network.fit(TRAINING_DATA, estimator=BayesianEstimator, prior_type="BDeu")
return network
def create_bayes_net():
atts = pd.read_csv('../../data/list_attr_celeba.csv')
atts = atts[KEEP_ATTS]
graph = BayesianModel()
graph.add_nodes_from(atts.columns)
graph.add_edges_from([('Young', 'Eyeglasses'), ('Young', 'Bald'),
('Young', 'Mustache'), ('Male', 'Mustache'),
('Male', 'Smiling'), ('Male', 'Wearing_Lipstick'),
('Young', 'Mouth_Slightly_Open'),
('Young', 'Narrow_Eyes'), ('Male', 'Narrow_Eyes'),
('Smiling', 'Narrow_Eyes'),
('Smiling', 'Mouth_Slightly_Open'),
('Young', 'Smiling')])
graph.fit(atts)
return graph
def generate_approx_model_from_graph(ebunch, nodes, df): """ Aprende un modelo Bayesiano de pgmpy usando un datos de un dataframe de pandas. Primero se hace un barajado de los datos. """ df = df.sample(frac=1) approx_model = BayesianModel(ebunch) approx_model.add_nodes_from(nodes) state_names = dict() for pair in ebunch: state_names[pair[0]] = [0, 1] state_names[pair[1]] = [0, 1] for node in nodes: state_names[node] = [0, 1] approx_model.fit(df, state_names=state_names, estimator=SmoothedMaximumLikelihoodEstimator) return approx_model
def bif2bayesian(pathname, verbose=3):
"""
Returns the fitted bayesian model
Example
----------
>>> from pgmpy.readwrite import BIFReader
>>> reader = BIFReader("bif_test.bif")
>>> reader.get_model()
<pgmpy.models.BayesianModel.BayesianModel object at 0x7f20af154320>
"""
if verbose >= 3: print('[BNLEARN] Loading bif file <%s>' % (pathname))
bifmodel = readwrite.BIF.BIFReader(path=pathname)
#bifmodel.get_edges()
try:
model = BayesianModel(bifmodel.variable_edges)
model.name = bifmodel.network_name
model.add_nodes_from(bifmodel.variable_names)
tabular_cpds = []
for var in sorted(bifmodel.variable_cpds.keys()):
values = bifmodel.variable_cpds[var]
cpd = TabularCPD(
var,
len(bifmodel.variable_states[var]),
values,
evidence=bifmodel.variable_parents[var],
evidence_card=[
len(bifmodel.variable_states[evidence_var])
for evidence_var in bifmodel.variable_parents[var]
])
tabular_cpds.append(cpd)
model.add_cpds(*tabular_cpds)
# for node, properties in bifmodel.variable_properties.items():
# for prop in properties:
# prop_name, prop_value = map(lambda t: t.strip(), prop.split('='))
# model.node[node][prop_name] = prop_value
return model
except AttributeError:
raise AttributeError(
'[BNLEARN] First get states of variables, edges, parents and network names'
)
def get_model(self):
"""
Returns an instance of Bayesian Model or Markov Model.
Varibles are in the pattern var_0, var_1, var_2 where var_0 is
0th index variable, var_1 is 1st index variable.
Return
------
model: an instance of Bayesian or Markov Model.
Examples
--------
>>> reader = UAIReader('TestUAI.uai')
>>> reader.get_model()
"""
if self.network_type == 'BAYES':
model = BayesianModel()
model.add_nodes_from(self.variables)
model.add_edges_from(self.edges)
tabular_cpds = []
for cpd in self.tables:
child_var = cpd[0]
states = int(self.domain[child_var])
arr = list(map(float, cpd[1]))
values = np.array(arr)
values = values.reshape(states, values.size // states)
tabular_cpds.append(TabularCPD(child_var, states, values))
model.add_cpds(*tabular_cpds)
return model
elif self.network_type == 'MARKOV':
model = MarkovModel(self.edges)
factors = []
for table in self.tables:
variables = table[0]
cardinality = [int(self.domain[var]) for var in variables]
value = list(map(float, table[1]))
factor = DiscreteFactor(variables=variables,
cardinality=cardinality,
values=value)
factors.append(factor)
model.add_factors(*factors)
return model
def createBayesGraph(graph_list,mapping,data):
'''
Creating the bayesian network graph and table
the graph_list, mapping and data are the parameters needed for creating the tables
this function returns:
bayes_model - the bayes model and its order
cpds_array - array of the tables
categories_each_element - categories of each element in the graph
'''
cpds_array = []
categories_each_element = {} # Returning an array with the values of each element
bayes_model = BayesianModel()
bayes_model.add_nodes_from(list(mapping))
for value in graph_list:
temp_list=value.split(',')
bayes_model.add_edge(temp_list[0],temp_list[1])
data_dict = {mapping[i]: data[:,i] for i in range(0, len(mapping))}
data_dict_pd = pandas.DataFrame(data=data_dict)
bayes_model.fit(data_dict_pd)
cpds_tables = bayes_model.get_cpds()
# Creating the array which returs to the client
for cpd in cpds_tables:
cpds_list = {}
for cat in cpd.state_names:
categories_each_element[cat] = cpd.state_names[cat]
cpd_string = str(cpd).split('|')
temp_array = []
cpd_matrix_values = []
digits_numbers = False
for a in cpd_string:
if (is_number(a)):
temp_array.append(float(a.strip()))
digits_numbers = True
elif ("-+" in a and digits_numbers == True):
cpd_matrix_values.append(temp_array)
temp_array = []
digits_numbers = False
cpds_list[str(list(cpd.variables))] = cpd_matrix_values
cpds_array.append(cpds_list)
return(bayes_model,cpds_array,categories_each_element)
def join(reference_bayes, second_bayes, new_dependent_vars,
new_independent_vars, ref_num_of_records, second_num_of_records):
final_bayes = BayesianModel()
#all independent variables should stay the same
final_bayes.add_nodes_from(new_independent_vars)
final_bayes.add_cpds(*[
reference_bayes.get_cpds(node=node) if node in
reference_bayes.nodes else second_bayes.get_cpds(node=node)
for node in new_independent_vars
])
for node in new_dependent_vars:
final_bayes.add_node(node)
ref_parents = set()
second_parents = set()
if node in reference_bayes:
ref_parents = set(reference_bayes.get_parents(node))
if node in second_bayes:
second_parents = set(second_bayes.get_parents(node))
if (len(ref_parents) == 0):
final_bayes.add_edges_from([(parent, node)
for parent in second_parents])
final_bayes.add_cpds(second_bayes.get_cpds(node=node))
else:
final_bayes.add_edges_from([(parent, node)
for parent in ref_parents])
if len(second_parents - ref_parents) > 0:
raise ValueError('This join can not be performed since the\
second distribution contains new independent variable\
(s) for node {}. Please consider dropping these new \
dependencies or switching reference distribution. '.
format(str(node)))
elif ref_parents == second_parents:
new_cpd = BayesNetHelper.calculate_weighted_cpds(
reference_bayes.get_cpds(node=node),
second_bayes.get_cpds(node=node), ref_num_of_records,
second_num_of_records)
final_bayes.add_cpds(new_cpd)
else:
final_bayes.add_cpds(reference_bayes.get_cpds(node=node))
return final_bayes
def create_bayes_net():
atts = pd.read_csv('./data/list_attr_celeba.csv')
atts = atts[KEEP_ATTS]
graph = BayesianModel()
graph.add_nodes_from(atts.columns)
# can't automate this part
# defining the structure of edges
graph.add_edges_from([('Young', 'Eyeglasses'), ('Young', 'Bald'),
('Young', 'Mustache'), ('Male', 'Mustache'),
('Male', 'Smiling'), ('Male', 'Wearing_Lipstick'),
('Young', 'Mouth_Slightly_Open'),
('Young', 'Narrow_Eyes'), ('Male', 'Narrow_Eyes'),
('Smiling', 'Narrow_Eyes'),
('Smiling', 'Mouth_Slightly_Open'),
('Young', 'Smiling')])
# fit estimates the CPD tables for the given structure
graph.fit(atts)
return graph
def create_network(models, processes, files):
for p in range(files):
temp_model = BayesianModel()
for e in range(len(processes[p].get_errors())):
temp_error = processes[p].get_error(e)
for c in range(len(temp_error.get_causes())):
temp_cause = temp_error.get_cause(c)
q = temp_cause.get_occ_prob(
) / temp_error.get_total_cause_prob()
temp_cause.set_occ_prob(q)
temp_model.add_nodes_from([temp_cause, temp_error])
temp_model.add_edge(temp_cause, temp_error)
temp_cause_cpd = TabularCPD(variable=temp_cause,
variable_card=2,
values=[[q, 1 - q]])
temp_model.add_cpds(temp_cause_cpd)
temp_error_cpd = TabularCPD(
variable=temp_error,
variable_card=2,
values=get_initial_error_cpd(len(temp_error.get_causes())),
evidence=temp_error.get_causes(),
evidence_card=[2] * (len(temp_error.get_causes())))
temp_model.add_cpds(temp_error_cpd)
for f in range(len(temp_error.get_effects())):
temp_effect = temp_error.get_effect(f)
temp_model.add_nodes_from([temp_error, temp_effect])
temp_model.add_edge(temp_error, temp_effect)
models.append(temp_model)
#plotting Failure Tree
dot = to_pydot(models[p])
with open('failure_tree_graph_%s.png' % processes[p], 'wb') as f:
f.write(dot.create_png())
#Sample output of CPDs for causes and errors
for e in range(len(processes[p].get_errors())):
for c in range(len(processes[p].get_error(e).get_causes())):
print(
temp_model.get_cpds(
processes[p].get_error(e).get_cause(c)))
print(temp_model.get_cpds(processes[p].get_error(e)))
def main():
data, string = readData()
genes = np.array(data.columns[1:])
labels = np.array(data.columns)
bayesianModel = BayesianModel()
transitionModel = DBN()
bayesianModel.add_nodes_from(genes)
transitionModel.add_nodes_from(genes)
bData, tData = getData(data, labels)
print "\nDynamic Bayesian Network inference",
print "\nB_0 network relations: "
hcb = HillClimbSearch(bData, genes, scoring_method=BicScore(bData, labels, bk1=string, weight=4))
best_model_b = hcb.estimate(start=bayesianModel, tabu_length=15, max_indegree=2)
print(best_model_b.edges())
printOutputB(best_model_b)
print "\nLocal Probability Model: "
best_model_b.fit(bData, BayesianEstimator)
for cpd in best_model_b.get_cpds():
print(cpd)
print "\nB_transition network relations: "
hct = HillClimbSearch(tData, genes, scoring_method=BicScore(tData, labels, bk1=string, weight=4))
best_model_t = hct.estimate_dynamic(start=transitionModel, tabu_length=15, max_indegree=2)
print(best_model_t.edges())
printOutputT(best_model_t)
print "\nLocal Probability Model: "
best_model_t.fit(tData, BayesianEstimator)
for cpd in best_model_t.get_cpds():
print(cpd)
def evaluate_single_graph(df_samples, graph, bn_truth, nb_repeat=3):
testing_graph = BayesianModel()
testing_graph.add_nodes_from(bn_truth.causal_graph.nodes())
for edge in remove_bidirected_edges(graph.edges()):
try:
testing_graph.add_edge(edge[0], edge[1])
except Exception as e:
try:
testing_graph.add_edge(edge[1], edge[0])
except Exception as e:
print(e)
continue
testing_graph.fit(df_samples, estimator=BayesianEstimator)
testing_graph.check_model()
bn_test = BayesianNetwork(testing_graph)
set_observe(bn_test.bn)
set_observe(bn_truth.bn)
bn_truth.set_state_names()
bn_test.set_state_names()
return {
'SID':
SID(bn_truth.causal_graph, bn_test.causal_graph),
'SHD':
SHD(bn_truth.causal_graph, bn_test.causal_graph),
'OD':
np.mean([
ODist(bn_truth, bn_test, 1000, discrete=True)
for _ in range(nb_repeat)
]),
'ID':
np.mean([
IDist(bn_truth, bn_test, 1000, discrete=True)
for _ in range(nb_repeat)
])
}
def reduce_model(self, evidence):
model = copy.deepcopy(self.model)
continuous_factors = [
factor for factor in model.factors
if isinstance(factor, ContinuousFactor)
]
for var, val in evidence.items():
for factor in continuous_factors:
if var in factor.scope(
) and "F(" in var: # make sure that we only reduce at this stage for continuous values, let the inference algorithm deal with reducing for binary variables
factor.reduce([(var, val)])
new_model = BayesianModel()
additional_evidence = {}
for node in model.factors:
if isinstance(node, ContinuousFactor):
if len(node.scope()) == 1:
node = TabularCPD(
str(node.scope()[0]), 2,
[[node.assignment(0),
node.assignment(1)]])
else:
node = to_CPD(node)
var = node.variable
for v in node.scope():
if var != v:
new_model.add_edge(str(v), str(var))
if "same_reason" in var:
additional_evidence[var] = 1
new_model.add_nodes_from([str(n) for n in node.scope()])
new_model.add_cpds(node)
return new_model, additional_evidence
def cal(self, file1, file2):
f1 = open(file1)
lines = f1.readlines()
nodes = self.getegdes(lines[0])
edges = self.getegdes(lines[1])
data = pd.read_csv(file2)
G = BayesianModel()
G.add_nodes_from(nodes)
for i in range(int(len(edges) / 2)):
G.add_edge(edges[2 * i], edges[2 * i + 1])
output1 = []
for i in range(int(len(edges) / 2)):
mut = mr.mutual_info_score(data[edges[2 * i]],
data[edges[2 * i + 1]])
output1.append(mut)
output2 = {}
for node1 in G.nodes():
d = {}
for node2 in G.nodes():
if node1 == node2:
continue
mut = mr.mutual_info_score(data[node1], data[node2])
d[node2] = mut
output2[node1] = d
print(output1)
print(output2)
with open('mutual_output.txt', 'w') as f:
f.write(str(output1))
f.write('\n')
f.write(str(output2))
class TestGibbsSampling(unittest.TestCase):
def setUp(self):
# A test Bayesian model
diff_cpd = TabularCPD('diff', 2, [[0.6], [0.4]])
intel_cpd = TabularCPD('intel', 2, [[0.7], [0.3]])
grade_cpd = TabularCPD('grade', 3, [[0.3, 0.05, 0.9, 0.5], [0.4, 0.25, 0.08, 0.3], [0.3, 0.7, 0.02, 0.2]],
evidence=['diff', 'intel'], evidence_card=[2, 2])
self.bayesian_model = BayesianModel()
self.bayesian_model.add_nodes_from(['diff', 'intel', 'grade'])
self.bayesian_model.add_edges_from([('diff', 'grade'), ('intel', 'grade')])
self.bayesian_model.add_cpds(diff_cpd, intel_cpd, grade_cpd)
# A test Markov model
self.markov_model = MarkovModel([('A', 'B'), ('C', 'B'), ('B', 'D')])
factor_ab = Factor(['A', 'B'], [2, 3], [1, 2, 3, 4, 5, 6])
factor_cb = Factor(['C', 'B'], [4, 3], [3, 1, 4, 5, 7, 8, 1, 3, 10, 4, 5, 6])
factor_bd = Factor(['B', 'D'], [3, 2], [5, 7, 2, 1, 9, 3])
self.markov_model.add_factors(factor_ab, factor_cb, factor_bd)
self.gibbs = GibbsSampling(self.bayesian_model)
def tearDown(self):
del self.bayesian_model
del self.markov_model
@patch('pgmpy.inference.Sampling.GibbsSampling._get_kernel_from_bayesian_model', autospec=True)
@patch('pgmpy.models.MarkovChain.__init__', autospec=True)
def test_init_bayesian_model(self, init, get_kernel):
model = MagicMock(spec_set=BayesianModel)
gibbs = GibbsSampling(model)
init.assert_called_once_with(gibbs)
get_kernel.assert_called_once_with(gibbs, model)
@patch('pgmpy.inference.Sampling.GibbsSampling._get_kernel_from_markov_model', autospec=True)
def test_init_markov_model(self, get_kernel):
model = MagicMock(spec_set=MarkovModel)
gibbs = GibbsSampling(model)
get_kernel.assert_called_once_with(gibbs, model)
def test_get_kernel_from_bayesian_model(self):
gibbs = GibbsSampling()
gibbs._get_kernel_from_bayesian_model(self.bayesian_model)
self.assertListEqual(list(gibbs.variables), self.bayesian_model.nodes())
self.assertDictEqual(gibbs.cardinalities, {'diff': 2, 'intel': 2, 'grade': 3})
def test_get_kernel_from_markov_model(self):
gibbs = GibbsSampling()
gibbs._get_kernel_from_markov_model(self.markov_model)
self.assertListEqual(list(gibbs.variables), self.markov_model.nodes())
self.assertDictEqual(gibbs.cardinalities, {'A': 2, 'B': 3, 'C': 4, 'D': 2})
def test_sample(self):
start_state = [State('diff', 0), State('intel', 0), State('grade', 0)]
sample = self.gibbs.sample(start_state, 2)
self.assertEquals(len(sample), 2)
self.assertEquals(len(sample.columns), 3)
self.assertIn('diff', sample.columns)
self.assertIn('intel', sample.columns)
self.assertIn('grade', sample.columns)
self.assertTrue(set(sample['diff']).issubset({0, 1}))
self.assertTrue(set(sample['intel']).issubset({0, 1}))
self.assertTrue(set(sample['grade']).issubset({0, 1, 2}))
@patch("pgmpy.inference.Sampling.GibbsSampling.random_state", autospec=True)
def test_sample_less_arg(self, random_state):
self.gibbs.state = None
random_state.return_value = [State('diff', 0), State('intel', 0), State('grade', 0)]
sample = self.gibbs.sample(size=2)
random_state.assert_called_once_with(self.gibbs)
self.assertEqual(len(sample), 2)
def test_generate_sample(self):
start_state = [State('diff', 0), State('intel', 0), State('grade', 0)]
gen = self.gibbs.generate_sample(start_state, 2)
samples = [sample for sample in gen]
self.assertEqual(len(samples), 2)
self.assertEqual({samples[0][0].var, samples[0][1].var, samples[0][2].var}, {'diff', 'intel', 'grade'})
self.assertEqual({samples[1][0].var, samples[1][1].var, samples[1][2].var}, {'diff', 'intel', 'grade'})
@patch("pgmpy.inference.Sampling.GibbsSampling.random_state", autospec=True)
def test_generate_sample_less_arg(self, random_state):
self.gibbs.state = None
gen = self.gibbs.generate_sample(size=2)
samples = [sample for sample in gen]
random_state.assert_called_once_with(self.gibbs)
self.assertEqual(len(samples), 2)
ax_temp.bar(x, z, zs=y, zdir='y', alpha=0.6, color='r' * 4)
ax_temp.set_xlabel('X')
ax_temp.set_ylabel('Y')
ax_temp.set_zlabel('Z')
ax_temp.title.set_text(('Feature ' + str(mean_indices[counter])))
counter += 1
plt.show()
# Learning naive bayes model from various subsets of data
naive_bayes_with_some_features(all_city_data, all_city_label, feature_list=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12])
naive_bayes_with_some_features(all_city_data, all_city_label, feature_list=[0, 1, 2])
naive_bayes_with_some_features(all_city_data, all_city_label, feature_list=[0, 1, 2, 4])
naive_bayes_with_some_features(all_city_data, all_city_label, feature_list=[0, 1, 2, 3, 4, 5])
# Splitting train and test data for PGM model
temp_data = pd.concat([all_city_data, pd.DataFrame(all_city_label, columns=[13])], axis=1)
pgm_train_set = temp_data.loc[0:700]
pgm_test_set = temp_data.loc[700:]
print(pgm_train_set)
# Implementing PGM model on data
# Using these features: 0: (age) 1: (sex) 2: (cp)
pgm_model = BayesianModel()
pgm_model.add_nodes_from([0, 1, 2, 13])
pgm_model.add_edges_from([(1, 13)])
pgm_model.fit(pgm_train_set.loc[:, [0, 1, 2, 13]])
pgm_test_set = pgm_test_set.loc[:, [0, 1, 2, 13]].drop(13, axis=1)
print(pgm_test_set)
print(pgm_model.get_cpds(13))
class TestBayesianModelFitPredict(unittest.TestCase):
def setUp(self):
self.model_disconnected = BayesianModel()
self.model_disconnected.add_nodes_from(['A', 'B', 'C', 'D', 'E'])
self.model_connected = BayesianModel([('A', 'B'), ('C', 'B'), ('C', 'D'), ('B', 'E')])
self.model2 = BayesianModel([('A', 'C'), ('B', 'C')])
self.data1 = pd.DataFrame(data={'A': [0, 0, 1], 'B': [0, 1, 0], 'C': [1, 1, 0]})
self.data2 = pd.DataFrame(data={'A': [0, np.NaN, 1],
'B': [0, 1, 0],
'C': [1, 1, np.NaN],
'D': [np.NaN, 'Y', np.NaN]})
# data_link - "https://www.kaggle.com/c/titanic/download/train.csv"
self.titanic_data = pd.read_csv('pgmpy/tests/test_estimators/testdata/titanic_train.csv', dtype=str)
self.titanic_data2 = self.titanic_data[["Survived", "Sex", "Pclass"]]
def test_bayesian_fit(self):
print(isinstance(BayesianEstimator, BaseEstimator))
print(isinstance(MaximumLikelihoodEstimator, BaseEstimator))
self.model2.fit(self.data1, estimator=BayesianEstimator, prior_type="dirichlet", pseudo_counts=[9, 3])
self.assertEqual(self.model2.get_cpds('B'), TabularCPD('B', 2, [[11.0 / 15], [4.0 / 15]]))
def test_fit_missing_data(self):
self.model2.fit(self.data2, state_names={'C': [0, 1]}, complete_samples_only=False)
cpds = set([TabularCPD('A', 2, [[0.5], [0.5]]),
TabularCPD('B', 2, [[2. / 3], [1. / 3]]),
TabularCPD('C', 2, [[0, 0.5, 0.5, 0.5], [1, 0.5, 0.5, 0.5]],
evidence=['A', 'B'], evidence_card=[2, 2])])
self.assertSetEqual(cpds, set(self.model2.get_cpds()))
def test_disconnected_fit(self):
values = pd.DataFrame(np.random.randint(low=0, high=2, size=(1000, 5)),
columns=['A', 'B', 'C', 'D', 'E'])
self.model_disconnected.fit(values)
for node in ['A', 'B', 'C', 'D', 'E']:
cpd = self.model_disconnected.get_cpds(node)
self.assertEqual(cpd.variable, node)
np_test.assert_array_equal(cpd.cardinality, np.array([2]))
value = (values.ix[:, node].value_counts() /
values.ix[:, node].value_counts().sum())
value = value.reindex(sorted(value.index)).values
np_test.assert_array_equal(cpd.values, value)
def test_predict(self):
titanic = BayesianModel()
titanic.add_edges_from([("Sex", "Survived"), ("Pclass", "Survived")])
titanic.fit(self.titanic_data2[500:])
p1 = titanic.predict(self.titanic_data2[["Sex", "Pclass"]][:30])
p2 = titanic.predict(self.titanic_data2[["Survived", "Pclass"]][:30])
p3 = titanic.predict(self.titanic_data2[["Survived", "Sex"]][:30])
p1_res = np.array(['0', '1', '0', '1', '0', '0', '0', '0', '0', '1', '0', '1', '0',
'0', '0', '1', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0',
'0', '0', '0', '0'])
p2_res = np.array(['male', 'female', 'female', 'female', 'male', 'male', 'male',
'male', 'female', 'female', 'female', 'female', 'male', 'male',
'male', 'female', 'male', 'female', 'male', 'female', 'male',
'female', 'female', 'female', 'male', 'female', 'male', 'male',
'female', 'male'])
p3_res = np.array(['3', '1', '1', '1', '3', '3', '3', '3', '1', '1', '1', '1', '3',
'3', '3', '1', '3', '1', '3', '1', '3', '1', '1', '1', '3', '1',
'3', '3', '1', '3'])
np_test.assert_array_equal(p1.values.ravel(), p1_res)
np_test.assert_array_equal(p2.values.ravel(), p2_res)
np_test.assert_array_equal(p3.values.ravel(), p3_res)
def test_connected_predict(self):
np.random.seed(42)
values = pd.DataFrame(np.array(np.random.randint(low=0, high=2, size=(1000, 5)),
dtype=str),
columns=['A', 'B', 'C', 'D', 'E'])
fit_data = values[:800]
predict_data = values[800:].copy()
self.model_connected.fit(fit_data)
self.assertRaises(ValueError, self.model_connected.predict, predict_data)
predict_data.drop('E', axis=1, inplace=True)
e_predict = self.model_connected.predict(predict_data)
np_test.assert_array_equal(e_predict.values.ravel(),
np.array([1, 0, 1, 0, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1,
1, 0, 0, 1, 1, 1, 0, 0, 0, 1, 1, 0, 1, 0,
0, 1, 0, 0, 0, 0, 1, 1, 1, 0, 1, 1, 0, 0,
0, 0, 0, 1, 1, 0, 1, 1, 0, 0, 0, 1, 0, 1,
0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 1,
1, 0, 1, 1, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1,
1, 1, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 0, 0,
1, 1, 1, 1, 0, 0, 1, 1, 0, 0, 0, 0, 1, 1,
0, 0, 0, 1, 0, 1, 1, 0, 1, 1, 0, 0, 1, 1,
1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1,
1, 0, 1, 0, 1, 1, 1, 1, 0, 0, 1, 0, 1, 1,
0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 1, 0, 1, 0,
1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 0, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 0, 1,
1, 1, 1, 0], dtype=str))
def test_connected_predict_probability(self):
np.random.seed(42)
values = pd.DataFrame(np.random.randint(low=0, high=2, size=(100, 5)),
columns=['A', 'B', 'C', 'D', 'E'])
fit_data = values[:80]
predict_data = values[80:].copy()
self.model_connected.fit(fit_data)
predict_data.drop('E', axis=1, inplace=True)
e_prob = self.model_connected.predict_probability(predict_data)
np_test.assert_allclose(e_prob.values.ravel(),
np.array([0.57894737, 0.42105263, 0.57894737, 0.42105263, 0.57894737,
0.42105263, 0.5 , 0.5 , 0.57894737, 0.42105263,
0.5 , 0.5 , 0.57894737, 0.42105263, 0.57894737,
0.42105263, 0.57894737, 0.42105263, 0.5 , 0.5 ,
0.57894737, 0.42105263, 0.57894737, 0.42105263, 0.5 ,
0.5 , 0.57894737, 0.42105263, 0.57894737, 0.42105263,
0.5 , 0.5 , 0.57894737, 0.42105263, 0.5 ,
0.5 , 0.5 , 0.5 , 0.5 , 0.5 ]), atol = 0)
predict_data = pd.DataFrame(np.random.randint(low=0, high=2, size=(1, 5)),
columns=['A', 'B', 'C', 'F', 'E'])[:]
def test_predict_probability_errors(self):
np.random.seed(42)
values = pd.DataFrame(np.random.randint(low=0, high=2, size=(2, 5)),
columns=['A', 'B', 'C', 'D', 'E'])
fit_data = values[:1]
predict_data = values[1:].copy()
self.model_connected.fit(fit_data)
self.assertRaises(ValueError, self.model_connected.predict_probability, predict_data)
predict_data = pd.DataFrame(np.random.randint(low=0, high=2, size=(1, 5)),
columns=['A', 'B', 'C', 'F', 'E'])[:]
self.assertRaises(ValueError, self.model_connected.predict_probability, predict_data)
def tearDown(self):
del self.model_connected
del self.model_disconnected
class TestBaseModelCreation(unittest.TestCase):
def setUp(self):
self.G = BayesianModel()
def test_class_init_without_data(self):
self.assertIsInstance(self.G, nx.DiGraph)
def test_class_init_with_data_string(self):
self.g = BayesianModel([('a', 'b'), ('b', 'c')])
self.assertListEqual(sorted(self.g.nodes()), ['a', 'b', 'c'])
self.assertListEqual(hf.recursive_sorted(self.g.edges()),
[['a', 'b'], ['b', 'c']])
def test_class_init_with_data_nonstring(self):
BayesianModel([(1, 2), (2, 3)])
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)
def test_add_nodes_from_string(self):
self.G.add_nodes_from(['a', 'b', 'c', 'd'])
self.assertListEqual(sorted(self.G.nodes()), ['a', 'b', 'c', 'd'])
def test_add_nodes_from_non_string(self):
self.G.add_nodes_from([1, 2, 3, 4])
def test_add_edge_string(self):
self.G.add_edge('d', 'e')
self.assertListEqual(sorted(self.G.nodes()), ['d', 'e'])
self.assertListEqual(self.G.edges(), [('d', 'e')])
self.G.add_nodes_from(['a', 'b', 'c'])
self.G.add_edge('a', 'b')
self.assertListEqual(hf.recursive_sorted(self.G.edges()),
[['a', 'b'], ['d', 'e']])
def test_add_edge_nonstring(self):
self.G.add_edge(1, 2)
def test_add_edge_selfloop(self):
self.assertRaises(ValueError, self.G.add_edge, 'a', 'a')
def test_add_edge_result_cycle(self):
self.G.add_edges_from([('a', 'b'), ('a', 'c')])
self.assertRaises(ValueError, self.G.add_edge, 'c', 'a')
def test_add_edges_from_string(self):
self.G.add_edges_from([('a', 'b'), ('b', 'c')])
self.assertListEqual(sorted(self.G.nodes()), ['a', 'b', 'c'])
self.assertListEqual(hf.recursive_sorted(self.G.edges()),
[['a', 'b'], ['b', 'c']])
self.G.add_nodes_from(['d', 'e', 'f'])
self.G.add_edges_from([('d', 'e'), ('e', 'f')])
self.assertListEqual(sorted(self.G.nodes()),
['a', 'b', 'c', 'd', 'e', 'f'])
self.assertListEqual(
hf.recursive_sorted(self.G.edges()),
hf.recursive_sorted([('a', 'b'), ('b', 'c'), ('d', 'e'),
('e', 'f')]))
def test_add_edges_from_nonstring(self):
self.G.add_edges_from([(1, 2), (2, 3)])
def test_add_edges_from_self_loop(self):
self.assertRaises(ValueError, self.G.add_edges_from, [('a', 'a')])
def test_add_edges_from_result_cycle(self):
self.assertRaises(ValueError, self.G.add_edges_from, [('a', 'b'),
('b', 'c'),
('c', 'a')])
def test_update_node_parents_bm_constructor(self):
self.g = BayesianModel([('a', 'b'), ('b', 'c')])
self.assertListEqual(self.g.predecessors('a'), [])
self.assertListEqual(self.g.predecessors('b'), ['a'])
self.assertListEqual(self.g.predecessors('c'), ['b'])
def test_update_node_parents(self):
self.G.add_nodes_from(['a', 'b', 'c'])
self.G.add_edges_from([('a', 'b'), ('b', 'c')])
self.assertListEqual(self.G.predecessors('a'), [])
self.assertListEqual(self.G.predecessors('b'), ['a'])
self.assertListEqual(self.G.predecessors('c'), ['b'])
def tearDown(self):
del self.G
class TestBaseEstimator(unittest.TestCase):
def setUp(self):
self.rand_data = pd.DataFrame(np.random.randint(0, 5, size=(5000, 2)), columns=list('AB'))
self.rand_data['C'] = self.rand_data['B']
self.est_rand = HillClimbSearch(self.rand_data, scoring_method=K2Score(self.rand_data))
self.model1 = BayesianModel()
self.model1.add_nodes_from(['A', 'B', 'C'])
self.model2 = self.model1.copy()
self.model2.add_edge('A', 'B')
# link to dataset: "https://www.kaggle.com/c/titanic/download/train.csv"
self.titanic_data = pd.read_csv('pgmpy/tests/test_estimators/testdata/titanic_train.csv')
self.titanic_data1 = self.titanic_data[["Survived", "Sex", "Pclass", "Age", "Embarked"]]
self.titanic_data2 = self.titanic_data[["Survived", "Sex", "Pclass"]]
self.est_titanic1 = HillClimbSearch(self.titanic_data1)
self.est_titanic2 = HillClimbSearch(self.titanic_data2)
def test_legal_operations(self):
model2_legal_ops = list(self.est_rand._legal_operations(self.model2))
model2_legal_ops_ref = [(('+', ('C', 'A')), -28.15602208305154),
(('+', ('A', 'C')), -28.155467430966382),
(('+', ('C', 'B')), 7636.947544933631),
(('+', ('B', 'C')), 7937.805375579936),
(('-', ('A', 'B')), 28.155467430966382),
(('flip', ('A', 'B')), -0.0005546520851567038)]
self.assertSetEqual(set([op for op, score in model2_legal_ops]),
set([op for op, score in model2_legal_ops_ref]))
def test_legal_operations_titanic(self):
est = self.est_titanic1
start_model = BayesianModel([("Survived", "Sex"),
("Pclass", "Age"),
("Pclass", "Embarked")])
legal_ops = est._legal_operations(start_model)
self.assertEqual(len(list(legal_ops)), 20)
tabu_list = [('-', ("Survived", "Sex")),
('-', ("Survived", "Pclass")),
('flip', ("Age", "Pclass"))]
legal_ops_tabu = est._legal_operations(start_model, tabu_list=tabu_list)
self.assertEqual(len(list(legal_ops_tabu)), 18)
legal_ops_indegree = est._legal_operations(start_model, max_indegree=1)
self.assertEqual(len(list(legal_ops_indegree)), 11)
legal_ops_both = est._legal_operations(start_model, tabu_list=tabu_list, max_indegree=1)
legal_ops_both_ref = [(('+', ('Embarked', 'Survived')), 10.050632580087608),
(('+', ('Survived', 'Pclass')), 41.88868046549101),
(('+', ('Age', 'Survived')), -23.635716036430836),
(('+', ('Pclass', 'Survived')), 41.81314459373226),
(('+', ('Sex', 'Pclass')), 4.772261678792802),
(('-', ('Pclass', 'Age')), 11.546515590731815),
(('-', ('Pclass', 'Embarked')), -32.171482832532774),
(('flip', ('Pclass', 'Embarked')), 3.3563814191281836),
(('flip', ('Survived', 'Sex')), 0.039737027979640516)]
self.assertSetEqual(set(legal_ops_both), set(legal_ops_both_ref))
def test_estimate_rand(self):
est1 = self.est_rand.estimate()
self.assertSetEqual(set(est1.nodes()), set(['A', 'B', 'C']))
self.assertTrue(est1.edges() == [('B', 'C')] or est1.edges() == [('C', 'B')])
est2 = self.est_rand.estimate(start=BayesianModel([('A', 'B'), ('A', 'C')]))
self.assertTrue(est2.edges() == [('B', 'C')] or est2.edges() == [('C', 'B')])
def test_estimate_titanic(self):
self.assertSetEqual(set(self.est_titanic2.estimate().edges()),
set([('Survived', 'Pclass'), ('Sex', 'Pclass'), ('Sex', 'Survived')]))
def tearDown(self):
del self.rand_data
del self.est_rand
del self.model1
del self.titanic_data
del self.titanic_data1
del self.titanic_data2
del self.est_titanic1
del self.est_titanic2
from pgmpy.models import BayesianModel
from pgmpy.factors import TabularCPD
# Creating the above bayesian network
model = BayesianModel()
model.add_nodes_from(['Rain', 'TrafficJam'])
model.add_edge('Rain', 'TrafficJam')
model.add_edge('Accident', 'TrafficJam')
cpd_rain = TabularCPD('Rain', 2, [[0.4], [0.6]])
cpd_accident = TabularCPD('Accident', 2, [[0.2], [0.8]])
cpd_traffic_jam = TabularCPD('TrafficJam', 2,
[[0.9, 0.6, 0.7, 0.1],
[0.1, 0.4, 0.3, 0.9]],
evidence=['Rain', 'Accident'],
evidence_card=[2, 2])
model.add_cpds(cpd_rain, cpd_accident, cpd_traffic_jam)
model.add_node('LongQueues')
model.add_edge('TrafficJam', 'LongQueues')
cpd_long_queues = TabularCPD('LongQueues', 2,
[[0.9, 0.2],
[0.1, 0.8]],
evidence=['TrafficJam'],
evidence_card=[2])
model.add_cpds(cpd_long_queues)
model.add_nodes_from(['GettingUpLate', 'LateForSchool'])
model.add_edges_from([('GettingUpLate', 'LateForSchool'),
('TrafficJam', 'LateForSchool')])
cpd_getting_up_late = TabularCPD('GettingUpLate', 2,
[[0.6], [0.4]])
cpd_late_for_school = TabularCPD('LateForSchool', 2,
[[0.9, 0.45, 0.8, 0.1],
[0.1, 0.55, 0.2, 0.9]],
class TestBaseModelCreation(unittest.TestCase):
def setUp(self):
self.G = BayesianModel()
def test_class_init_without_data(self):
self.assertIsInstance(self.G, nx.DiGraph)
def test_class_init_with_data_string(self):
self.g = BayesianModel([('a', 'b'), ('b', 'c')])
self.assertListEqual(sorted(self.g.nodes()), ['a', 'b', 'c'])
self.assertListEqual(hf.recursive_sorted(self.g.edges()),
[['a', 'b'], ['b', 'c']])
def test_class_init_with_data_nonstring(self):
BayesianModel([(1, 2), (2, 3)])
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)
def test_add_nodes_from_string(self):
self.G.add_nodes_from(['a', 'b', 'c', 'd'])
self.assertListEqual(sorted(self.G.nodes()), ['a', 'b', 'c', 'd'])
def test_add_nodes_from_non_string(self):
self.G.add_nodes_from([1, 2, 3, 4])
def test_add_edge_string(self):
self.G.add_edge('d', 'e')
self.assertListEqual(sorted(self.G.nodes()), ['d', 'e'])
self.assertListEqual(self.G.edges(), [('d', 'e')])
self.G.add_nodes_from(['a', 'b', 'c'])
self.G.add_edge('a', 'b')
self.assertListEqual(hf.recursive_sorted(self.G.edges()),
[['a', 'b'], ['d', 'e']])
def test_add_edge_nonstring(self):
self.G.add_edge(1, 2)
def test_add_edge_selfloop(self):
self.assertRaises(ValueError, self.G.add_edge, 'a', 'a')
def test_add_edge_result_cycle(self):
self.G.add_edges_from([('a', 'b'), ('a', 'c')])
self.assertRaises(ValueError, self.G.add_edge, 'c', 'a')
def test_add_edges_from_string(self):
self.G.add_edges_from([('a', 'b'), ('b', 'c')])
self.assertListEqual(sorted(self.G.nodes()), ['a', 'b', 'c'])
self.assertListEqual(hf.recursive_sorted(self.G.edges()),
[['a', 'b'], ['b', 'c']])
self.G.add_nodes_from(['d', 'e', 'f'])
self.G.add_edges_from([('d', 'e'), ('e', 'f')])
self.assertListEqual(sorted(self.G.nodes()),
['a', 'b', 'c', 'd', 'e', 'f'])
self.assertListEqual(hf.recursive_sorted(self.G.edges()),
hf.recursive_sorted([('a', 'b'), ('b', 'c'),
('d', 'e'), ('e', 'f')]))
def test_add_edges_from_nonstring(self):
self.G.add_edges_from([(1, 2), (2, 3)])
def test_add_edges_from_self_loop(self):
self.assertRaises(ValueError, self.G.add_edges_from,
[('a', 'a')])
def test_add_edges_from_result_cycle(self):
self.assertRaises(ValueError, self.G.add_edges_from,
[('a', 'b'), ('b', 'c'), ('c', 'a')])
def test_update_node_parents_bm_constructor(self):
self.g = BayesianModel([('a', 'b'), ('b', 'c')])
self.assertListEqual(self.g.predecessors('a'), [])
self.assertListEqual(self.g.predecessors('b'), ['a'])
self.assertListEqual(self.g.predecessors('c'), ['b'])
def test_update_node_parents(self):
self.G.add_nodes_from(['a', 'b', 'c'])
self.G.add_edges_from([('a', 'b'), ('b', 'c')])
self.assertListEqual(self.G.predecessors('a'), [])
self.assertListEqual(self.G.predecessors('b'), ['a'])
self.assertListEqual(self.G.predecessors('c'), ['b'])
def tearDown(self):
del self.G
def pdag_to_dag(pdag):
"""Completes a PDAG to a DAG, without adding v-structures, if such a
completion exists. If no faithful extension is possible, some fully
oriented DAG that corresponds to the PDAG is returned and a warning is
generated. This is a static method.
Parameters
----------
pdag: DirectedGraph
A directed acyclic graph pattern, consisting in (acyclic) directed edges
as well as "undirected" edges, represented as both-way edges between
nodes.
Returns
-------
dag: BayesianModel
A faithful orientation of pdag, if one exists. Otherwise any
fully orientated DAG/BayesianModel with the structure of pdag.
References
----------
[1] Chickering, Learning Equivalence Classes of Bayesian-Network Structures,
2002; See page 454 (last paragraph) for the algorithm pdag_to_dag
http://www.jmlr.org/papers/volume2/chickering02a/chickering02a.pdf
[2] Dor & Tarsi, A simple algorithm to construct a consistent extension
of a partially oriented graph, 1992,
http://ftp.cs.ucla.edu/pub/stat_ser/r185-dor-tarsi.pdf
Examples
--------
>>> import pandas as pd
>>> import numpy as np
>>> from pgmpy.base import DirectedGraph
>>> from pgmpy.estimators import ConstraintBasedEstimator
>>> data = pd.DataFrame(np.random.randint(0, 4, size=(5000, 3)), columns=list('ABD'))
>>> data['C'] = data['A'] - data['B']
>>> data['D'] += data['A']
>>> c = ConstraintBasedEstimator(data)
>>> pdag = c.skeleton_to_pdag(*c.estimate_skeleton())
>>> pdag.edges()
[('B', 'C'), ('D', 'A'), ('A', 'D'), ('A', 'C')]
>>> c.pdag_to_dag(pdag).edges()
[('B', 'C'), ('A', 'D'), ('A', 'C')]
>>> # pdag_to_dag is static:
... pdag1 = DirectedGraph([('A', 'B'), ('C', 'B'), ('C', 'D'), ('D', 'C'), ('D', 'A'), ('A', 'D')])
>>> ConstraintBasedEstimator.pdag_to_dag(pdag1).edges()
[('D', 'C'), ('C', 'B'), ('A', 'B'), ('A', 'D')]
>>> # example of a pdag with no faithful extension:
... pdag2 = DirectedGraph([('A', 'B'), ('A', 'C'), ('B', 'C'), ('C', 'B')])
>>> ConstraintBasedEstimator.pdag_to_dag(pdag2).edges()
UserWarning: PDAG has no faithful extension (= no oriented DAG with the same v-structures as PDAG).
Remaining undirected PDAG edges oriented arbitrarily.
[('B', 'C'), ('A', 'B'), ('A', 'C')]
"""
pdag = pdag.copy()
dag = BayesianModel()
dag.add_nodes_from(pdag.nodes())
# add already directed edges of pdag to dag
for X, Y in pdag.edges():
if not pdag.has_edge(Y, X):
dag.add_edge(X, Y)
while pdag.number_of_nodes() > 0:
# find node with (1) no directed outgoing edges and
# (2) the set of undirected neighbors is either empty or
# undirected neighbors + parents of X are a clique
found = False
for X in pdag.nodes():
directed_outgoing_edges = set(pdag.successors(X)) - set(pdag.predecessors(X))
undirected_neighbors = set(pdag.successors(X)) & set(pdag.predecessors(X))
neighbors_are_clique = all((pdag.has_edge(Y, Z)
for Z in pdag.predecessors(X)
for Y in undirected_neighbors if not Y == Z))
if not directed_outgoing_edges and \
(not undirected_neighbors or neighbors_are_clique):
found = True
# add all edges of X as outgoing edges to dag
for Y in pdag.predecessors(X):
dag.add_edge(Y, X)
pdag.remove_node(X)
break
if not found:
warn("PDAG has no faithful extension (= no oriented DAG with the " +
"same v-structures as PDAG). Remaining undirected PDAG edges " +
"oriented arbitrarily.")
for X, Y in pdag.edges():
if not dag.has_edge(Y, X):
try:
dag.add_edge(X, Y)
except ValueError:
pass
break
return dag
# Bayesian network for students
from pgmpy.models import BayesianModel
model = BayesianModel()
# Add nodes
model.add_nodes_from(['difficulty', 'intelligence', 'grade', 'sat', 'letter'])
print(model.nodes())
# Add edges
model.add_edges_from([('difficulty', 'grade'), ('intelligence', 'grade'), ('intelligence', 'sat'), ('grade', 'letter')])
print(model.edges())
