def load_total_cpds():
# All the nodes in the graph (157 nodes)
gnodes = total_G.nodes
data = pd.DataFrame(np.random.randint(low=0,
high=2,
size=(100, len(gnodes))),
columns=gnodes)
# Option 1 of fitting cpds
estimator = BayesianEstimator(total_G, data)
p = estimator.get_parameters(prior_type='BDeu', equivalent_sample_size=5)
for i, cpd in enumerate(p):
total_G.add_cpds(cpd)
# Option 2 of fitting cpds
for i in range(1, num_sub_symptoms + 1):
cpd_sub = estimator.estimate_cpd('sub_sympt_' + str(i),
prior_type="BDeu")
total_G.add_cpds(cpd_sub)
if i <= num_symptoms:
cpd_symp = estimator.estimate_cpd('sympt_' + str(i),
prior_type="BDeu")
total_G.add_cpds(cpd_symp)
# this is the time cruncher.
for i in range(1, num_conditions + 1):
cpd_cond = estimator.estimate_cpd('cond_' + str(i), prior_type="BDeu")
total_G.add_cpds(cpd_cond)
class TimeBayesianEstimator:
timeout = 1200
def setup(self):
self.alarm = get_example_model('alarm')
self.alarm_model = BayesianNetwork(self.alarm.edges())
self.alarm_est = BayesianEstimator(
self.alarm_model, self.alarm.simulate(int(1e4),
show_progress=False))
self.munin = get_example_model('munin1')
self.munin_model = BayesianNetwork(self.munin.edges())
self.munin_est = BayesianEstimator(
self.munin_model, self.munin.simulate(int(1e4),
show_progress=False))
def time_alarm_bayesian_estimator(self):
self.alarm_est.get_parameters()
def time_munin_bayesian_estimator(self):
self.munin_est.get_parameters()
mydata = np.genfromtxt(
r'E:\Lessons_tutorials\Behavioural user profile articles\Datasets\7 twor.2009\twor.2009\converted\pgmpy\activities+time_ordered_withoutdatetime.csv',
delimiter=",")
#pd.read_csv(r'E:\Lessons_tutorials\Behavioural user profile articles\Datasets\7 twor.2009\twor.2009\converted\pgmpy\data.csv')
#print(mydata)
data = pd.DataFrame(mydata, columns=feature_names) #['X', 'Y'])
print(data)
list_of_scoring_methods = [
BicScore(data),
#BdeuScore(data),
#K2Score(data)
]
for scoreMethod in list_of_scoring_methods:
start_time = time.time()
hc = HillClimbSearch(data, scoreMethod)
best_model = hc.estimate()
print(hc.scoring_method)
print(best_model.edges())
end_time = time.time()
print("execution time in seconds:")
print(end_time - start_time)
estimator = BayesianEstimator(best_model, data)
print(estimator.get_parameters(prior_type='K2')) #, equivalent_sample_size=5)
#casas7_model = BayesianModel()
#casas7_model.fit(data, estimator=BayesianEstimator)#MaximumLikelihoodEstimator)
#print(casas7_model.get_cpds())
#casas7_model.get_n
best_model = est.estimate()
print("Structure found!")
print("Best Model")
print(best_model.edges())
'''
Parameters Estimation and CPD tables
'''
from pgmpy.models import BayesianModel
from pgmpy.estimators import BayesianEstimator
model = BayesianModel(best_model.edges())
data = pd.read_csv("final_training_set.csv")
estimator = BayesianEstimator(model, data)
parameters = estimator.get_parameters(prior_type='BDeu',
equivalent_sample_size=10)
for cpd in parameters:
model.add_cpds(cpd)
'''
Inference and Validation
'''
from pgmpy.inference import VariableElimination
import csv
f = open("validation_data.csv")
reader = csv.reader(f)
inference = VariableElimination(model)
valid = 0
invalid = 0
path = r'C:\Users\Javier\Documents\MEGA\Universitattt\Master\Thesis\CDS_data\toy_data'
file1 = r'epsilon_csv.csv'
data = pd.read_csv(os.path.join(path,file1), index_col = 0, skiprows = [19 + i for i in range(2)]).transpose()
print('Using BicScore Method')
est = HillClimbSearch(data, scoring_method = BicScore(data))
best_model = est.estimate()
# MLE_estimator = MaximumLikelihoodEstimator(best_model, data)
# MLE_parameters = MLE_estimator.get_parameters()
bay_estimator = BayesianEstimator(best_model, data)
bay_parameters = bay_estimator.get_parameters()
print('Edges:')
print(best_model.edges())
# print('MLE Parameters')
# for m in MLE_parameters:
# print(m)
print('Bayesian Parameters')
for b in bay_parameters:
print(b)
class TestBayesianEstimator(unittest.TestCase):
def setUp(self):
self.m1 = BayesianModel([('A', 'C'), ('B', 'C')])
self.d1 = pd.DataFrame(data={'A': [0, 0, 1], 'B': [0, 1, 0], 'C': [1, 1, 0]})
self.d2 = pd.DataFrame(data={'A': [0, 0, 1, 0, 2, 0, 2, 1, 0, 2],
'B': ['X', 'Y', 'X', 'Y', 'X', 'Y', 'X', 'Y', 'X', 'Y'],
'C': [1, 1, 1, 0, 0, 0, 0, 0, 0, 0]})
self.est1 = BayesianEstimator(self.m1, self.d1)
self.est2 = BayesianEstimator(self.m1, self.d1, state_names={'A': [0, 1, 2],
'B': [0, 1],
'C': [0, 1, 23]})
self.est3 = BayesianEstimator(self.m1, self.d2)
def test_estimate_cpd_dirichlet(self):
cpd_A = self.est1.estimate_cpd('A', prior_type="dirichlet", pseudo_counts=[0, 1])
self.assertEqual(cpd_A, TabularCPD('A', 2, [[0.5], [0.5]]))
cpd_B = self.est1.estimate_cpd('B', prior_type="dirichlet", pseudo_counts=[9, 3])
self.assertEqual(cpd_B, TabularCPD('B', 2, [[11.0/15], [4.0/15]]))
cpd_C = self.est1.estimate_cpd('C', prior_type="dirichlet", pseudo_counts=[0.4, 0.6])
self.assertEqual(cpd_C, TabularCPD('C', 2, [[0.2, 0.2, 0.7, 0.4],
[0.8, 0.8, 0.3, 0.6]],
evidence=['A', 'B'], evidence_card=[2, 2]))
def test_estimate_cpd_improper_prior(self):
cpd_C = self.est1.estimate_cpd('C', prior_type="dirichlet", pseudo_counts=[0, 0])
cpd_C_correct = (TabularCPD('C', 2, [[0.0, 0.0, 1.0, np.NaN],
[1.0, 1.0, 0.0, np.NaN]],
evidence=['A', 'B'], evidence_card=[2, 2],
state_names={'A': [0, 1], 'B': [0, 1], 'C': [0, 1]}))
# manual comparison because np.NaN != np.NaN
self.assertTrue(((cpd_C.values == cpd_C_correct.values) |
np.isnan(cpd_C.values) & np.isnan(cpd_C_correct.values)).all())
def test_estimate_cpd_shortcuts(self):
cpd_C1 = self.est2.estimate_cpd('C', prior_type='BDeu', equivalent_sample_size=9)
cpd_C1_correct = TabularCPD('C', 3, [[0.2, 0.2, 0.6, 1./3, 1./3, 1./3],
[0.6, 0.6, 0.2, 1./3, 1./3, 1./3],
[0.2, 0.2, 0.2, 1./3, 1./3, 1./3]],
evidence=['A', 'B'], evidence_card=[3, 2])
self.assertEqual(cpd_C1, cpd_C1_correct)
cpd_C2 = self.est3.estimate_cpd('C', prior_type='K2')
cpd_C2_correct = TabularCPD('C', 2, [[0.5, 0.6, 1./3, 2./3, 0.75, 2./3],
[0.5, 0.4, 2./3, 1./3, 0.25, 1./3]],
evidence=['A', 'B'], evidence_card=[3, 2])
self.assertEqual(cpd_C2, cpd_C2_correct)
def test_get_parameters(self):
cpds = set([self.est3.estimate_cpd('A'),
self.est3.estimate_cpd('B'),
self.est3.estimate_cpd('C')])
self.assertSetEqual(set(self.est3.get_parameters()), cpds)
def test_get_parameters2(self):
pseudo_counts = {'A': [1, 2, 3], 'B': [4, 5], 'C': [6, 7]}
cpds = set([self.est3.estimate_cpd('A', prior_type="dirichlet", pseudo_counts=pseudo_counts['A']),
self.est3.estimate_cpd('B', prior_type="dirichlet", pseudo_counts=pseudo_counts['B']),
self.est3.estimate_cpd('C', prior_type="dirichlet", pseudo_counts=pseudo_counts['C'])])
self.assertSetEqual(set(self.est3.get_parameters(prior_type="dirichlet",
pseudo_counts=pseudo_counts)), cpds)
def tearDown(self):
del self.m1
del self.d1
del self.d2
del self.est1
del self.est2
]
prior_type = ["dirichlet", "BDeu", "K2"]
for i in range(1, 3):
f = open(
r'C:\Users\Adele\Desktop\PhD thesis tests\parameterLearning_' +
prior_type[i], 'w')
print(prior_type[i])
model = BayesianModel(structures[i])
estimator = BayesianEstimator(model, data)
end_time = time.time()
print("parameter learning in seconds:{}".format(end_time - start_time))
all_cpds = estimator.get_parameters(
prior_type=prior_type[i]) #BDeu")#"dirichlet")
for c in all_cpds:
print(c)
f.write(c.tostring()) # python will convert \n to os.linesep
f.close()
#for my_node in feature_names:
# the_cpd = estimator.estimate_cpd(my_node, prior_type="dirichlet")
# print(the_cpd)
#casas7_model = BayesianModel()
#casas7_model.fit(data, estimator=BayesianEstimator)#MaximumLikelihoodEstimator)
#print(casas7_model.get_cpds())
#casas7_model.get_n
print(mle.estimate_cpd(node='CVP'))
# Estimating CPDs for all the nodes in the model
print(mle.get_parameters()[:10]) # Show just the first 10 CPDs in the output
# Verifying that the learned parameters are almost equal.
import numpy as np
print(np.allclose(alarm_model.get_cpds('FIO2').values, mle.estimate_cpd('FIO2').values, atol=0.01))
# Fitting the using Bayesian Estimator
from pgmpy.estimators import BayesianEstimator
best = BayesianEstimator(model=model_struct, data=samples)
print(best.estimate_cpd(node='FIO2', prior_type="BDeu", equivalent_sample_size=1000))
# Uniform pseudo count for each state. Can also accept an array of the size of CPD.
print(best.estimate_cpd(node='CVP', prior_type="dirichlet", pseudo_counts=100))
# Learning CPDs for all the nodes in the model. For learning all parameters with BDeU prior, a dict of
# pseudo_counts need to be provided
print(best.get_parameters(prior_type="BDeu", equivalent_sample_size=1000)[:10])
# Shortcut for learning all the parameters and adding the CPDs to the model.
model_struct = BayesianModel(ebunch=alarm_model.edges())
model_struct.fit(data=samples, estimator=MaximumLikelihoodEstimator)
print(model_struct.get_cpds('FIO2'))
model_struct = BayesianModel(ebunch=alarm_model.edges())
model_struct.fit(data=samples, estimator=BayesianEstimator, prior_type='BDeu', equivalent_sample_size=1000)
print(model_struct.get_cpds('FIO2'))
class TestBayesianEstimator(unittest.TestCase):
def setUp(self):
self.m1 = BayesianModel([('A', 'C'), ('B', 'C')])
self.d1 = pd.DataFrame(data={
'A': [0, 0, 1],
'B': [0, 1, 0],
'C': [1, 1, 0]
})
self.d2 = pd.DataFrame(
data={
'A': [0, 0, 1, 0, 2, 0, 2, 1, 0, 2],
'B': ['X', 'Y', 'X', 'Y', 'X', 'Y', 'X', 'Y', 'X', 'Y'],
'C': [1, 1, 1, 0, 0, 0, 0, 0, 0, 0]
})
self.est1 = BayesianEstimator(self.m1, self.d1)
self.est2 = BayesianEstimator(self.m1,
self.d1,
state_names={
'A': [0, 1, 2],
'B': [0, 1],
'C': [0, 1, 23]
})
self.est3 = BayesianEstimator(self.m1, self.d2)
def test_estimate_cpd_dirichlet(self):
cpd_A = self.est1.estimate_cpd('A',
prior_type="dirichlet",
pseudo_counts=[[0], [1]])
self.assertEqual(cpd_A, TabularCPD('A', 2, [[0.5], [0.5]]))
cpd_B = self.est1.estimate_cpd('B',
prior_type="dirichlet",
pseudo_counts=[[9], [3]])
self.assertEqual(cpd_B, TabularCPD('B', 2, [[11.0 / 15], [4.0 / 15]]))
cpd_C = self.est1.estimate_cpd('C',
prior_type="dirichlet",
pseudo_counts=[[0.4, 0.4, 0.4, 0.4],
[0.6, 0.6, 0.6, 0.6]])
self.assertEqual(
cpd_C,
TabularCPD('C',
2, [[0.2, 0.2, 0.7, 0.4], [0.8, 0.8, 0.3, 0.6]],
evidence=['A', 'B'],
evidence_card=[2, 2]))
def test_estimate_cpd_improper_prior(self):
cpd_C = self.est1.estimate_cpd('C',
prior_type="dirichlet",
pseudo_counts=[[0, 0, 0, 0],
[0, 0, 0, 0]])
cpd_C_correct = (TabularCPD(
'C',
2, [[0.0, 0.0, 1.0, np.NaN], [1.0, 1.0, 0.0, np.NaN]],
evidence=['A', 'B'],
evidence_card=[2, 2],
state_names={
'A': [0, 1],
'B': [0, 1],
'C': [0, 1]
}))
# manual comparison because np.NaN != np.NaN
self.assertTrue(
((cpd_C.values == cpd_C_correct.values)
| np.isnan(cpd_C.values) & np.isnan(cpd_C_correct.values)).all())
def test_estimate_cpd_shortcuts(self):
cpd_C1 = self.est2.estimate_cpd('C',
prior_type='BDeu',
equivalent_sample_size=9)
cpd_C1_correct = TabularCPD('C',
3,
[[0.2, 0.2, 0.6, 1. / 3, 1. / 3, 1. / 3],
[0.6, 0.6, 0.2, 1. / 3, 1. / 3, 1. / 3],
[0.2, 0.2, 0.2, 1. / 3, 1. / 3, 1. / 3]],
evidence=['A', 'B'],
evidence_card=[3, 2])
self.assertEqual(cpd_C1, cpd_C1_correct)
cpd_C2 = self.est3.estimate_cpd('C', prior_type='K2')
cpd_C2_correct = TabularCPD('C',
2,
[[0.5, 0.6, 1. / 3, 2. / 3, 0.75, 2. / 3],
[0.5, 0.4, 2. / 3, 1. / 3, 0.25, 1. / 3]],
evidence=['A', 'B'],
evidence_card=[3, 2])
self.assertEqual(cpd_C2, cpd_C2_correct)
def test_get_parameters(self):
cpds = set([
self.est3.estimate_cpd('A'),
self.est3.estimate_cpd('B'),
self.est3.estimate_cpd('C')
])
self.assertSetEqual(set(self.est3.get_parameters()), cpds)
def test_get_parameters2(self):
pseudo_counts = {
'A': [[1], [2], [3]],
'B': [[4], [5]],
'C': [[6, 6, 6, 6, 6, 6], [7, 7, 7, 7, 7, 7]]
}
cpds = set([
self.est3.estimate_cpd('A',
prior_type="dirichlet",
pseudo_counts=pseudo_counts['A']),
self.est3.estimate_cpd('B',
prior_type="dirichlet",
pseudo_counts=pseudo_counts['B']),
self.est3.estimate_cpd('C',
prior_type="dirichlet",
pseudo_counts=pseudo_counts['C'])
])
self.assertSetEqual(
set(
self.est3.get_parameters(prior_type="dirichlet",
pseudo_counts=pseudo_counts)), cpds)
def tearDown(self):
del self.m1
del self.d1
del self.d2
del self.est1
del self.est2
class TestBayesianEstimator(unittest.TestCase):
def setUp(self):
self.m1 = BayesianModel([("A", "C"), ("B", "C")])
self.d1 = pd.DataFrame(data={"A": [0, 0, 1], "B": [0, 1, 0], "C": [1, 1, 0]})
self.d2 = pd.DataFrame(
data={
"A": [0, 0, 1, 0, 2, 0, 2, 1, 0, 2],
"B": ["X", "Y", "X", "Y", "X", "Y", "X", "Y", "X", "Y"],
"C": [1, 1, 1, 0, 0, 0, 0, 0, 0, 0],
}
)
self.est1 = BayesianEstimator(self.m1, self.d1)
self.est2 = BayesianEstimator(
self.m1, self.d1, state_names={"A": [0, 1, 2], "B": [0, 1], "C": [0, 1, 23]}
)
self.est3 = BayesianEstimator(self.m1, self.d2)
def test_estimate_cpd_dirichlet(self):
cpd_A = self.est1.estimate_cpd(
"A", prior_type="dirichlet", pseudo_counts=[[0], [1]]
)
self.assertEqual(cpd_A, TabularCPD("A", 2, [[0.5], [0.5]]))
cpd_B = self.est1.estimate_cpd(
"B", prior_type="dirichlet", pseudo_counts=[[9], [3]]
)
self.assertEqual(cpd_B, TabularCPD("B", 2, [[11.0 / 15], [4.0 / 15]]))
cpd_C = self.est1.estimate_cpd(
"C",
prior_type="dirichlet",
pseudo_counts=[[0.4, 0.4, 0.4, 0.4], [0.6, 0.6, 0.6, 0.6]],
)
self.assertEqual(
cpd_C,
TabularCPD(
"C",
2,
[[0.2, 0.2, 0.7, 0.4], [0.8, 0.8, 0.3, 0.6]],
evidence=["A", "B"],
evidence_card=[2, 2],
),
)
def test_estimate_cpd_improper_prior(self):
cpd_C = self.est1.estimate_cpd(
"C", prior_type="dirichlet", pseudo_counts=[[0, 0, 0, 0], [0, 0, 0, 0]]
)
cpd_C_correct = TabularCPD(
"C",
2,
[[0.0, 0.0, 1.0, np.NaN], [1.0, 1.0, 0.0, np.NaN]],
evidence=["A", "B"],
evidence_card=[2, 2],
state_names={"A": [0, 1], "B": [0, 1], "C": [0, 1]},
)
# manual comparison because np.NaN != np.NaN
self.assertTrue(
(
(cpd_C.values == cpd_C_correct.values)
| np.isnan(cpd_C.values) & np.isnan(cpd_C_correct.values)
).all()
)
def test_estimate_cpd_shortcuts(self):
cpd_C1 = self.est2.estimate_cpd(
"C", prior_type="BDeu", equivalent_sample_size=9
)
cpd_C1_correct = TabularCPD(
"C",
3,
[
[0.2, 0.2, 0.6, 1.0 / 3, 1.0 / 3, 1.0 / 3],
[0.6, 0.6, 0.2, 1.0 / 3, 1.0 / 3, 1.0 / 3],
[0.2, 0.2, 0.2, 1.0 / 3, 1.0 / 3, 1.0 / 3],
],
evidence=["A", "B"],
evidence_card=[3, 2],
)
self.assertEqual(cpd_C1, cpd_C1_correct)
cpd_C2 = self.est3.estimate_cpd("C", prior_type="K2")
cpd_C2_correct = TabularCPD(
"C",
2,
[
[0.5, 0.6, 1.0 / 3, 2.0 / 3, 0.75, 2.0 / 3],
[0.5, 0.4, 2.0 / 3, 1.0 / 3, 0.25, 1.0 / 3],
],
evidence=["A", "B"],
evidence_card=[3, 2],
)
self.assertEqual(cpd_C2, cpd_C2_correct)
def test_get_parameters(self):
cpds = set(
[
self.est3.estimate_cpd("A"),
self.est3.estimate_cpd("B"),
self.est3.estimate_cpd("C"),
]
)
self.assertSetEqual(set(self.est3.get_parameters()), cpds)
def test_get_parameters2(self):
pseudo_counts = {
"A": [[1], [2], [3]],
"B": [[4], [5]],
"C": [[6, 6, 6, 6, 6, 6], [7, 7, 7, 7, 7, 7]],
}
cpds = set(
[
self.est3.estimate_cpd(
"A", prior_type="dirichlet", pseudo_counts=pseudo_counts["A"]
),
self.est3.estimate_cpd(
"B", prior_type="dirichlet", pseudo_counts=pseudo_counts["B"]
),
self.est3.estimate_cpd(
"C", prior_type="dirichlet", pseudo_counts=pseudo_counts["C"]
),
]
)
self.assertSetEqual(
set(
self.est3.get_parameters(
prior_type="dirichlet", pseudo_counts=pseudo_counts
)
),
cpds,
)
def tearDown(self):
del self.m1
del self.d1
del self.d2
del self.est1
del self.est2