def testget_paramsFail(self):
# Get params when not all features/class have been enumerated
dataPath = os.path.join(THIS_DIR, "test/DataSets/Tests/StringData.csv")
se = StringEnumerator(dataPath, "phenotype")
with self.assertRaises(Exception) as context:
se.get_params()
self.assertTrue("Features and Phenotypes must be fully numeric" in str(
context.exception))
def testContValues5000Iterations(self):
dataPath = os.path.join(THIS_DIR, "test/DataSets/Real/ContinuousAndNonBinaryDiscreteAttributes.csv")
converter = StringEnumerator(dataPath,"Class")
headers, classLabel, dataFeatures, dataPhenotypes = converter.get_params()
clf = XCS(learning_iterations=5000)
clf.fit(dataFeatures,dataPhenotypes)
answer = 0.64
#print("Continuous Attributes 5000 Iter: "+str(clf.get_final_training_accuracy()))
self.assertTrue(self.approxEqualOrBetter(0.2,clf.get_final_training_accuracy(),answer,True))
def test20BitMP5000Iterations(self):
dataPath = os.path.join(THIS_DIR, "test/DataSets/Real/Multiplexer20Modified.csv")
converter = StringEnumerator(dataPath,"Class")
headers, classLabel, dataFeatures, dataPhenotypes = converter.get_params()
clf = XCS(learning_iterations=5000,N=2000,nu=10)
clf.fit(dataFeatures,dataPhenotypes)
answer = 0.6634
#print("20 Bit 5000 Iter: "+str(clf.get_final_training_accuracy()))
self.assertTrue(self.approxEqualOrBetter(0.2,clf.get_final_training_accuracy(),answer,True))
def testPredictInvVar(self):
dataPath = os.path.join(THIS_DIR,
"test/DataSets/Real/Multiplexer6Modified.csv")
converter = StringEnumerator(dataPath, "Class")
headers, classLabel, dataFeatures, dataPhenotypes = converter.get_params(
)
clf = XCS(learning_iterations=1000,
N=500,
nu=10,
use_inverse_varinance=True,
p_explore=0.5)
clf.fit(dataFeatures, dataPhenotypes)
print("kkkkkkkkkkkkkkkkkkkkkkkkkkk")
print(clf.predict(clf.env.formatData.savedRawTrainingData[0]))
def testInverseVariance(self):
dataPath = os.path.join(THIS_DIR,
"test/DataSets/Real/Multiplexer11.csv")
converter = StringEnumerator(dataPath, "class")
headers, classLabel, dataFeatures, dataPhenotypes = converter.get_params(
)
clf = XCS(learning_iterations=5000,
N=1000,
mixing_method="inv-var-only-mixing")
clf.fit(dataFeatures, dataPhenotypes)
answer = 0.894
score = clf.get_final_training_accuracy()
print("#####################################\n6 Bit 1000 Iter: " +
str(score))
def testContValuesAndMissingTesting5000Iterations(self):
dataPath = os.path.join(THIS_DIR, "test/DataSets/Real/ContinuousAndNonBinaryDiscreteAttributesMissing.csv")
converter = StringEnumerator(dataPath, "Class")
headers, classLabel, dataFeatures, dataPhenotypes = converter.get_params()
formatted = np.insert(dataFeatures, dataFeatures.shape[1], dataPhenotypes, 1)
np.random.shuffle(formatted)
dataFeatures = np.delete(formatted, -1, axis=1)
dataPhenotypes = formatted[:, -1]
clf = XCS(learning_iterations=5000)
score = np.mean(cross_val_score(clf, dataFeatures, dataPhenotypes, cv=3))
answer = 0.5
#print("Cont & Missing Testing 5000 Iter: " + str(score))
self.assertTrue(self.approxEqualOrBetter(0.2, score, answer, True))
def test6BitMPTesting1000Iterations(self):
dataPath = os.path.join(THIS_DIR, "test/DataSets/Real/Multiplexer6Modified.csv")
converter = StringEnumerator(dataPath,"Class")
headers, classLabel, dataFeatures, dataPhenotypes = converter.get_params()
formatted = np.insert(dataFeatures, dataFeatures.shape[1], dataPhenotypes, 1)
np.random.shuffle(formatted)
dataFeatures = np.delete(formatted, -1, axis=1)
dataPhenotypes = formatted[:, -1]
clf = XCS(learning_iterations=1000,N=500,nu=10)
score = np.mean(cross_val_score(clf, dataFeatures, dataPhenotypes, cv=3))
answer = 0.9
#print("6 Bit Testing 1000 Iter: "+str(score))
self.assertTrue(self.approxEqualOrBetter(0.2,score,answer,True))
def testNew(self):
#Use StringEnumerator to gather data
converter = StringEnumerator("test/DataSets/Real/Multiplexer11.csv",
"class")
headers, actionLabel, dataFeatures, dataActions = converter.get_params(
)
#Shuffle data
formatted = np.insert(dataFeatures, dataFeatures.shape[1], dataActions,
1)
np.random.shuffle(formatted)
dataFeatures = np.delete(formatted, -1, axis=1)
dataActions = formatted[:, -1]
#Initialize and train model
clf_inv_var = XCS(learning_iterations=1000,
N=200,
use_inverse_varinance=True)
clf_inv_var.fit(dataFeatures, dataActions)
breakpoint()
def testget_params2(self):
# Get Params Test
dataPath = os.path.join(THIS_DIR, "test/DataSets/Tests/StringData.csv")
se = StringEnumerator(dataPath, "phenotype")
se.change_header_name("N1", "gender")
se.change_header_name("N2", "floats")
se.change_header_name("N3", "age")
se.change_class_name("country")
se.add_attribute_converter("gender",
np.array(["female", "male", "NA", "other"]))
se.add_attribute_converter("age", np.array(["old", "young"]))
se.add_class_converter_random()
se.convert_all_attributes()
dataHeaders, classLabel, dataFeatures, dataPhenotypes = se.get_params()
cHeaders = np.array(["gender", "floats", "age"])
cFeatures = np.array([[1, 1.2, 1], [0, 0.3, np.nan], [0, -0.4, 0],
[np.nan, 0, 1]])
cPhenotypes = np.array([0, 1, 0, 2])
self.assertEqual("country", classLabel)
self.assertTrue(np.array_equal(cHeaders, dataHeaders))
self.assertTrue(np.allclose(cFeatures, dataFeatures, equal_nan=True))
self.assertTrue(
np.allclose(cPhenotypes, dataPhenotypes, equal_nan=True))