def test_build_classifier(self):
"""
Tests the build_classifier method.
"""
# 1. nominal
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
data.class_is_last()
cname = "weka.classifiers.trees.J48"
options = ["-C", "0.3"]
cls = classifiers.Classifier(classname=cname, options=options)
self.assertIsNotNone(cls, msg="Failed to instantiate: " + cname + "/" + str(options))
cls.build_classifier(data)
# 2. numeric
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("bolts.arff"))
self.assertIsNotNone(data)
data.class_is_last()
cname = "weka.classifiers.functions.LinearRegression"
options = ["-R", "0.1"]
cls = classifiers.Classifier(classname=cname, options=options)
self.assertIsNotNone(cls, msg="Failed to instantiate: " + cname + "/" + str(options))
cls.build_classifier(data)
def test_classify_instance(self):
"""
Tests the classify_instance method.
"""
# 1. nominal
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
data.class_is_last()
cname = "weka.classifiers.trees.J48"
options = ["-C", "0.3"]
cls = classifiers.Classifier(classname=cname, options=options)
self.assertIsNotNone(cls,
msg="Failed to instantiate: " + cname + "/" +
str(options))
cls.build_classifier(data)
preds = []
for i in range(10, 20):
pred = cls.classify_instance(data.get_instance(i))
self.assertIsNotNone(pred)
preds.append(pred)
self.assertEqual([2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 5.0, 5.0, 2.0, 2.0],
preds,
msg="Classifications differ")
# 2. numeric
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("bolts.arff"))
self.assertIsNotNone(data)
data.class_is_last()
cname = "weka.classifiers.functions.LinearRegression"
options = ["-R", "0.1"]
cls = classifiers.Classifier(classname=cname, options=options)
self.assertIsNotNone(cls,
msg="Failed to instantiate: " + cname + "/" +
str(options))
cls.build_classifier(data)
preds = []
for i in range(10):
pred = cls.classify_instance(data.get_instance(i))
self.assertIsNotNone(pred)
preds.append(pred)
expected = [
24.313, 33.359, 28.569, 26.365, 32.680, 29.149, 26.998, 22.971,
13.160, 7.394
]
for i in range(len(preds)):
self.assertAlmostEqual(expected[i],
preds[i],
places=3,
msg="Classifications differ")
def test_distribution_for_instance(self):
"""
Tests the distribution_for_instance method.
"""
# 1. nominal
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
data.class_is_last()
cname = "weka.classifiers.trees.J48"
options = ["-C", "0.3"]
cls = classifiers.Classifier(classname=cname, options=options)
self.assertIsNotNone(cls,
msg="Failed to instantiate: " + cname + "/" +
str(options))
cls.build_classifier(data)
for i in range(10):
dist = cls.distribution_for_instance(data.get_instance(i))
self.assertIsNotNone(dist)
self.assertEqual(6,
len(dist),
msg="Number of classes in prediction differ!")
# 2. numeric
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("bolts.arff"))
self.assertIsNotNone(data)
data.class_is_last()
cname = "weka.classifiers.functions.LinearRegression"
options = ["-R", "0.1"]
cls = classifiers.Classifier(classname=cname, options=options)
self.assertIsNotNone(cls,
msg="Failed to instantiate: " + cname + "/" +
str(options))
cls.build_classifier(data)
for i in range(10):
dist = cls.distribution_for_instance(data.get_instance(i))
self.assertIsNotNone(dist)
self.assertEqual(
1,
len(dist),
msg=
"Number of classes in prediction should be one for numeric classifier!"
)
def test_generate_thresholdcurve_data(self):
"""
Tests the generate_thresholdcurve_data method.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("diabetes.arff"))
data.class_is_last()
remove = filters.Filter(
classname="weka.filters.unsupervised.attribute.Remove",
options=["-R", "1-3"])
cls = classifiers.Classifier(
classname="weka.classifiers.bayes.NaiveBayes")
fc = classifiers.FilteredClassifier()
fc.filter = remove
fc.classifier = cls
evl = classifiers.Evaluation(data)
evl.crossvalidate_model(cls, data, 10, Random(1))
data = plot.generate_thresholdcurve_data(evl, 0)
self.assertEqual(13,
data.num_attributes,
msg="number of attributes differs")
self.assertEqual(769, data.num_instances, msg="number of rows differs")
attname = "True Positives"
self.assertIsNotNone(data.attribute_by_name(attname),
msg="Failed to locate attribute: " + attname)
attname = "False Positive Rate"
self.assertIsNotNone(data.attribute_by_name(attname),
msg="Failed to locate attribute: " + attname)
attname = "Lift"
self.assertIsNotNone(data.attribute_by_name(attname),
msg="Failed to locate attribute: " + attname)
def test_stats(self):
"""
Tests the Stats class.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data, msg="Failed to load data!")
stats = data.attribute_stats(3)
numstats = stats.numeric_stats
self.assertAlmostEqual(898,
numstats.count,
places=3,
msg="count differs")
self.assertAlmostEqual(70, numstats.max, places=3, msg="max differs")
self.assertAlmostEqual(3.635,
numstats.mean,
places=3,
msg="mean differs")
self.assertAlmostEqual(0.0, numstats.min, places=3, msg="min differs")
self.assertAlmostEqual(13.717,
numstats.stddev,
places=3,
msg="stddev differs")
self.assertAlmostEqual(3264, numstats.sum, places=3, msg="sum differs")
self.assertAlmostEqual(180636,
numstats.sumsq,
places=3,
msg="sumsq differs")
def test_build_clusterer(self):
"""
Tests the build_clusterer class.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
data.delete_last_attribute()
# batch
cls = clusterers.Clusterer(classname="weka.clusterers.SimpleKMeans")
self.assertIsNotNone(cls, msg="Failed to instantiate clusterer!")
cls.build_clusterer(data)
self.assertIsNotNone(
str(cls), msg="Failed to obtain string representation of model")
# incremental
cls = clusterers.Clusterer(classname="weka.clusterers.Cobweb")
self.assertIsNotNone(cls, msg="Failed to instantiate clusterer!")
cls.build_clusterer(dataset.Instances.template_instances(data))
for inst in data:
cls.update_clusterer(inst)
cls.update_finished()
self.assertIsNotNone(
str(cls), msg="Failed to obtain string representation of model")
def test_attribute_selection(self):
"""
Tests attribute selection.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
data.class_is_last()
cname = "weka.attributeSelection.BestFirst"
options = ["-D", "1", "-N", "5"]
search = attribute_selection.ASSearch(classname=cname, options=options)
self.assertIsNotNone(search, msg="Search should not be None: " + cname + "/" + str(options))
cname = "weka.attributeSelection.CfsSubsetEval"
options = ["-P", "1", "-E", "1"]
evaluation = attribute_selection.ASEvaluation(classname=cname, options=options)
self.assertIsNotNone(evaluation, msg="Evaluation should not be None: " + cname + "/" + str(options))
attsel = attribute_selection.AttributeSelection()
self.assertIsNotNone(search, msg="AttributeSelection should not be None!")
attsel.search(search)
attsel.evaluator(evaluation)
attsel.select_attributes(data)
self.assertEqual(9, attsel.number_attributes_selected, msg="number_attributes_selected differs")
self.assertEqual([0, 4, 8, 11, 12, 19, 24, 32, 33, 38], attsel.selected_attributes.tolist(), msg="selected_attributes differ")
self.assertGreater(len(attsel.results_string), 0, msg="results_string should get produced")
def test_build_and_use_forecaster(self):
"""
Tests building and using of a forecaster.
"""
self._ensure_package_is_installed()
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("airline.arff"))
self.assertIsNotNone(data, msg="Data should not be none")
data.class_is_last()
airline_train, airline_test = data.train_test_split(90.0)
forecaster = timeseries.WekaForecaster()
self.assertIsNotNone(forecaster)
forecaster.fields_to_forecast = ["passenger_numbers"]
forecaster.base_forecaster = classifiers.Classifier(
classname="weka.classifiers.functions.LinearRegression")
forecaster.fields_to_forecast = "passenger_numbers"
forecaster.build_forecaster(airline_train)
num_prime_instances = 12
airline_prime = dataset.Instances.copy_instances(
airline_train, airline_train.num_instances - num_prime_instances,
num_prime_instances)
forecaster.prime_forecaster(airline_prime)
num_future_forecasts = airline_test.num_instances
preds = forecaster.forecast(num_future_forecasts)
self.assertIsNotNone(preds, msg="Predictions should not be none")
self.assertEqual(len(preds),
airline_test.num_instances,
msg="# of predictions should equal prime window size")
def test_batch_filtering(self):
"""
Tests the Filter.filter method.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
flter = filters.Filter(
classname="weka.filters.unsupervised.attribute.Remove",
options=["-R", "1,3"])
flter.inputformat(data)
filtered = flter.filter(data)
self.assertEqual(data.num_attributes - 2,
filtered.num_attributes,
msg="Number of attributes differ")
self.assertEqual(data.num_instances,
filtered.num_instances,
msg="Number of instances differ")
# multple files
data = loader.load_file(
self.datafile("reutersTop10Randomized_1perc_shortened-train.arff"))
self.assertIsNotNone(data)
data2 = loader.load_file(
self.datafile("reutersTop10Randomized_1perc_shortened-test.arff"))
self.assertIsNotNone(data2)
flter = filters.Filter(
classname="weka.filters.unsupervised.attribute.StringToWordVector")
flter.inputformat(data)
filtered = flter.filter([data, data2])
self.assertIsNone(filtered[0].equal_headers(filtered[1]),
msg="Headers should be compatible")
def test_classes_to_clusters(self):
"""
Tests the classes_to_clusters method.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
#data.class_is_last()
train = dataset.Instances.copy_instances(data, 0, data.num_instances)
train.delete_last_attribute()
cls = clusterers.Clusterer(classname="weka.clusterers.SimpleKMeans")
self.assertIsNotNone(cls, msg="Failed to instantiate clusterer!")
cls.build_clusterer(train)
test = dataset.Instances.copy_instances(data, 0, data.num_instances)
test.class_is_last()
evl = clusterers.ClusterEvaluation()
self.assertIsNotNone(cls, msg="Failed to instantiate evaluation!")
evl.set_model(cls)
evl.test_model(test)
self.assertEqual([2, 4],
evl.classes_to_clusters.tolist(),
msg="classes to clusters differs")
def test_reduce(self):
"""
Tests reducing of attributes.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
data.class_is_last()
cname = "weka.attributeSelection.BestFirst"
options = ["-D", "1", "-N", "5"]
search = attribute_selection.ASSearch(classname=cname, options=options)
self.assertIsNotNone(search, msg="Search should not be None: " + cname + "/" + str(options))
cname = "weka.attributeSelection.CfsSubsetEval"
options = ["-P", "1", "-E", "1"]
evaluation = attribute_selection.ASEvaluation(classname=cname, options=options)
self.assertIsNotNone(evaluation, msg="Evaluation should not be None: " + cname + "/" + str(options))
attsel = attribute_selection.AttributeSelection()
self.assertIsNotNone(search, msg="AttributeSelection should not be None!")
attsel.search(search)
attsel.evaluator(evaluation)
attsel.select_attributes(data)
reduced = attsel.reduce_dimensionality(data)
self.assertEqual(attsel.number_attributes_selected + 1, reduced.num_attributes, msg="number of attributes differs")
self.assertEqual(data.num_instances, reduced.num_instances, msg="number of rows differs")
def test_ranking(self):
"""
Tests ranking of attributes.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
data.class_is_last()
search = attribute_selection.ASSearch(
classname="weka.attributeSelection.Ranker", options=["-N", "-1"])
self.assertIsNotNone(search, msg="Search should not be None!")
evaluation = attribute_selection.ASEvaluation(
"weka.attributeSelection.InfoGainAttributeEval")
self.assertIsNotNone(evaluation, msg="Evaluation should not be None!")
attsel = attribute_selection.AttributeSelection()
self.assertIsNotNone(attsel,
msg="AttributeSelection should not be None!")
attsel.ranking(True)
attsel.folds(2)
attsel.crossvalidation(True)
attsel.seed(42)
attsel.search(search)
attsel.evaluator(evaluation)
attsel.select_attributes(data)
self.assertGreater(len(str(attsel.ranked_attributes)),
0,
msg="results_string should get produced")
self.assertGreater(len(attsel.results_string),
0,
msg="results_string should get produced")
def test_distribution_for_instance(self):
"""
Tests the distribution_for_instance method.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
data.delete_last_attribute()
cls = clusterers.Clusterer(classname="weka.clusterers.SimpleKMeans")
self.assertIsNotNone(cls, msg="Failed to instantiate clusterer!")
cls.build_clusterer(data)
preds = []
for i in range(10):
dist = cls.distribution_for_instance(data.get_instance(i))
self.assertIsNotNone(
dist, msg="Failed to obtain cluster membership for instance")
self.assertEqual(2, len(dist), msg="Number of clusters differs")
preds.append(dist)
expected = [[0., 1.], [1., 0.], [1., 0.], [0., 1.], [0., 1.], [0., 1.],
[1., 0.], [1., 0.], [1., 0.], [1., 0.]]
self.assertEqual(
len(expected),
len(preds),
msg="Expected/predicted differ in length - update required!")
for i in range(len(expected)):
self.assertEqual(expected[i],
preds[i].tolist(),
msg="Cluster distributions differ")
def load_Arff_file(file):
if type(file) is not str:
file = str(file)
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(file)
return data
def test_clusterevaluation(self):
"""
Tests the ClusterEvaluation class.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
data.delete_last_attribute()
# simple test set
test = dataset.Instances.copy_instances(data, 0, 10)
cls = clusterers.Clusterer(classname="weka.clusterers.SimpleKMeans")
self.assertIsNotNone(cls, msg="Failed to instantiate clusterer!")
cls.build_clusterer(data)
evl = clusterers.ClusterEvaluation()
self.assertIsNotNone(cls, msg="Failed to instantiate evaluation!")
evl.set_model(cls)
evl.test_model(test)
self.assertGreater(len(evl.cluster_results), 0, msg="No evaluation string!")
self.assertEqual(0.0, evl.log_likelihood, msg="log_likelihood differs")
self.assertEqual(2, evl.num_clusters, msg="num_clusters differs")
self.assertEquals(
[1., 0., 0., 1., 1., 1., 0., 0., 0., 0.], evl.cluster_assignments.tolist(),
msg="cluster_assignments differs")
# cross-validation
cls = clusterers.Clusterer(classname="weka.clusterers.EM", options=["-I", "3", "-X", "2", "-max", "5"])
self.assertIsNotNone(cls, msg="Failed to instantiate clusterer!")
llh = clusterers.ClusterEvaluation.crossvalidate_model(cls, data, 10, classes.Random(1))
self.assertAlmostEqual(-34.397, llh, places=3, msg="Failed to cross-validate clusterer!")
def test_batchpredictor(self):
"""
Tests the batch predictor methods.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
data.class_is_last()
cname = "weka.classifiers.trees.J48"
options = []
cls = classifiers.Classifier(classname=cname, options=options)
self.assertIsNotNone(cls, msg="Failed to instantiate: " + cname + "/" + str(options))
# batch predictor?
self.assertTrue(cls.is_batchpredictor, msg="not a batch predictor: " + cname + "/" + str(options))
# more efficient implementation?
cls.has_efficient_batch_prediction()
# batch size
self.assertIsNotNone(cls.batch_size, msg="batch size is not initialized")
# distributions_for_instances
cls.build_classifier(data)
dists = cls.distributions_for_instances(data)
self.assertIsNotNone(dists, msg="no distributions generated")
self.assertEqual(len(dists), len(data), msg="number of predictions differ")
self.assertEqual(len(dists[0]), data.class_attribute.num_values, msg="size of distribution array does not match number of classes")
def test_incremental_filtering(self):
"""
Tests the Filter.input/output methods.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
flter = filters.Filter(
classname="weka.filters.unsupervised.attribute.Remove",
options=["-R", "1,3"])
flter.inputformat(data)
filtered = flter.outputformat()
self.assertIsNotNone(filtered)
self.assertTrue(isinstance(filtered, dataset.Instances),
msg="Should be Instances object")
for inst in data:
flter.input(inst)
finst = flter.output()
filtered.add_instance(finst)
self.assertEqual(data.num_attributes - 2,
filtered.num_attributes,
msg="Number of attributes differ")
self.assertEqual(data.num_instances,
filtered.num_instances,
msg="Number of instances differ")
def test_matrix_plot(self):
"""
Tests the matrix_plot method.
"""
loader = converters.Loader("weka.core.converters.ArffLoader")
iris_data = loader.load_file(self.datafile("iris.arff"))
iris_data.class_is_last()
plot.matrix_plot(iris_data, percent=50, title="Matrix plot iris", wait=False)
def test_instance(self):
"""
Tests the Instance class.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.ORIG.arff"))
self.assertIsNotNone(data, msg="Failed to load data!")
inst = data.get_instance(0)
self.assertEqual(39, inst.num_attributes, msg="num_attributes differs")
self.assertEqual(-1, data.class_index, msg="class_index differs")
data.class_index = data.num_attributes - 1
self.assertEqual(38, data.class_index, msg="class_index differs")
data.class_is_first()
self.assertEqual(0, data.class_index, msg="class_index differs")
data.class_is_last()
self.assertEqual(38, data.class_index, msg="class_index differs")
self.assertIsNotNone(inst.dataset, msg="Dataset reference should not be None!")
self.assertTrue(inst.has_missing(), msg="Should have missing values")
self.assertTrue(inst.is_missing(0), msg="First value should be missing")
self.assertFalse(inst.is_missing(1), msg="Second value should not be missing")
self.assertEqual("C", inst.get_string_value(1), msg="string value differs")
inst.set_string_value(1, "H")
self.assertEqual("H", inst.get_string_value(1), msg="string value differs")
self.assertEqual(8, inst.get_value(3), msg="numeric value differs")
inst.set_value(3, 6.3)
self.assertEqual(6.3, inst.get_value(3), msg="numeric value differs")
self.assertEqual(1, inst.weight, msg="weight should be 1")
inst.weight = 0.5
self.assertEqual(0.5, inst.weight, msg="weights differ")
values = [1.0, 2.0, 3.0]
inst = dataset.Instance.create_instance(values, classname="weka.core.DenseInstance")
self.assertEqual(3, inst.num_attributes, msg="#attributes differ")
self.assertEqual(1.0, inst.get_value(0), msg="value at #" + str(0) + " differs")
self.assertEqual(2.0, inst.get_value(1), msg="value at #" + str(1) + " differs")
self.assertEqual(3.0, inst.get_value(2), msg="value at #" + str(2) + " differs")
values = [0.0, 2.0, 0.0]
inst = dataset.Instance.create_instance(values, classname="weka.core.SparseInstance")
self.assertEqual(3, inst.num_attributes, msg="#attributes differ")
self.assertEqual(0.0, inst.get_value(0), msg="value at #" + str(0) + " differs")
self.assertEqual(2.0, inst.get_value(1), msg="value at #" + str(1) + " differs")
self.assertEqual(0.0, inst.get_value(2), msg="value at #" + str(2) + " differs")
values = [(1, 2.0)]
inst = dataset.Instance.create_sparse_instance(values, 3, classname="weka.core.SparseInstance")
self.assertEqual(3, inst.num_attributes, msg="#attributes differ")
self.assertEqual(0.0, inst.get_value(0), msg="value at #" + str(0) + " differs")
self.assertEqual(2.0, inst.get_value(1), msg="value at #" + str(1) + " differs")
self.assertEqual(0.0, inst.get_value(2), msg="value at #" + str(2) + " differs")
def test_arff_loader(self):
"""
Tests the Loader class using an ArffLoader.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
self.assertEqual(898,
data.num_instances,
msg="Number of instances differs!")
def test_incremental_arff_loader(self):
"""
Tests the Loader class using an incremental ArffLoader.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"), incremental=True)
self.assertIsNotNone(data)
count = 0
for inst in loader:
count += 1
self.assertEqual(898, count, msg="Number of instances differs!")
def test_scatter_plot(self):
"""
Tests the scatter_plot method.
"""
loader = converters.Loader("weka.core.converters.ArffLoader")
iris_data = loader.load_file(self.datafile("iris.arff"))
iris_data.class_is_last()
plot.scatter_plot(iris_data,
iris_data.attribute_by_name("petalwidth").index,
iris_data.attribute_by_name("petallength").index,
percent=50,
wait=False)
def test_learning_curve(self):
"""
Tests the learning_curve method.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("diabetes.arff"))
data.class_is_last()
cls = [
classifiers.Classifier(classname="weka.classifiers.trees.J48"),
classifiers.Classifier(classname="weka.classifiers.bayes.NaiveBayesUpdateable")]
plot.plot_learning_curve(cls, data, wait=False)
plot.plot_learning_curve(cls, data, increments=0.1, wait=False)
def test_plot_classifier_errors(self):
"""
Tests the plot_classifier_errors method.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
bolts_data = loader.load_file(self.datafile("bolts.arff"))
self.assertIsNotNone(bolts_data)
bolts_data.class_is_last()
classifier = classifiers.Classifier(classname="weka.classifiers.functions.LinearRegression", options=["-S", "1", "-C"])
evaluation = classifiers.Evaluation(bolts_data)
evaluation.crossvalidate_model(classifier, bolts_data, 10, Random(42))
plot.plot_classifier_errors(evaluation.predictions, wait=False)
def test_plot_dot_graph(self):
"""
Tests the plot_dot_graph method.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("diabetes.arff"))
data.class_is_last()
cls = classifiers.Classifier(classname="weka.classifiers.trees.J48",
options=["-C", "0.3"])
cls.build_classifier(data)
graph.plot_dot_graph(cls.graph)
def test_arff_saver(self):
"""
Tests the Saver class using an ArffSaver.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data)
outfile = self.tempfile("out.arff")
self.delfile(outfile)
saver = converters.Saver(classname="weka.core.converters.ArffSaver")
saver.save_file(data, outfile)
self.assertTrue(os.path.exists(outfile),
"File does not exist: " + outfile)
self.delfile(outfile)
def test_build_associator(self):
"""
Tests the build_classifier method.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("nursery.arff"))
self.assertIsNotNone(data)
data.class_is_last()
cname = "weka.associations.Apriori"
options = None
cls = associations.Associator(classname=cname, options=options)
self.assertIsNotNone(cls,
msg="Failed to instantiate: " + cname + "/" +
str(options))
cls.build_associations(data)
def test_plot_prc(self):
"""
Tests the plot_prc method.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("diabetes.arff"))
data.class_is_last()
remove = filters.Filter(classname="weka.filters.unsupervised.attribute.Remove", options=["-R", "1-3"])
cls = classifiers.Classifier(classname="weka.classifiers.bayes.NaiveBayes")
fc = classifiers.FilteredClassifier()
fc.filter = remove
fc.classifier = cls
evl = classifiers.Evaluation(data)
evl.crossvalidate_model(cls, data, 10, Random(1))
plot.plot_prc(evl, class_index=[0, 1], wait=False)
def test_attributestats(self):
"""
Tests the AttributeStats class.
"""
loader = converters.Loader(classname="weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("anneal.arff"))
self.assertIsNotNone(data, msg="Failed to load data!")
stats = data.attribute_stats(2)
self.assertIsNotNone(stats, msg="Failed to obtain stats!")
self.assertEqual(8, stats.distinct_count, "distinct_count differs")
self.assertEqual(898, stats.int_count, "int_count differs")
self.assertEqual(0, stats.missing_count, "missing_count differs")
self.assertEqual([86, 256, 440, 0, 51, 20, 10, 19, 16], stats.nominal_counts.tolist(), "nominal_counts differs")
self.assertEqual([86, 256, 440, 0, 51, 20, 10, 19, 16], stats.nominal_weights.tolist(), "nominal_weights differs")
self.assertEqual(898, stats.total_count, "total_count differs")
self.assertEqual(0, stats.unique_count, "unique_count differs")
def test_plot_cluster_assignments(self):
"""
Tests the plot_cluster_assignments method.
"""
loader = converters.Loader("weka.core.converters.ArffLoader")
data = loader.load_file(self.datafile("iris.arff"))
data.delete_last_attribute()
# build a clusterer and output model
clusterer = clusterers.Clusterer(
classname="weka.clusterers.SimpleKMeans", options=["-N", "3"])
clusterer.build_clusterer(data)
evaluation = clusterers.ClusterEvaluation()
evaluation.set_model(clusterer)
evaluation.test_model(data)
plot.plot_cluster_assignments(evaluation,
data,
inst_no=True,
wait=False)
