def testInitFileImporter(self):
#no filename
with self.assertRaises(TypeError):
LibsvmFileImporter()
#wrong filename
with self.assertRaises(IOError):
LibsvmFileImporter('efwef')
def testBinaryImport(self):
cwd = os.path.dirname(os.path.abspath(sys.argv[0]))
# should do
importer = LibsvmFileImporter(os.path.join(cwd,'data/classification/a1a'),binary=True)
importer.get_dataSet()
# should fail
with self.assertRaises(TypeError): #multi-class data
importer = LibsvmFileImporter(os.path.join(cwd,'data/classification/satimage.scale'),binary=True)
def validateModel(self, testFile):
testdata = LibsvmFileImporter(testFile).get_dataSet()
self.__inst_test = testdata.get_numInstances()
## --- statistics
correct = 0.
sum_error = 0
for i in testdata.get_targets():
if i == 1: #correct
correct += 1.
else:
sum_error += math.pow(1 - i, 2)
# percent correct
self.__pct_correct = 100 * (correct / self.__inst_test)
# root mean squared error
self.__rmse = math.sqrt(sum_error / self.__inst_test)
def validateModel(self, testFile):
testdata = LibsvmFileImporter(testFile).get_dataSet()
self.__inst_test = testdata.get_numInstances()
## --- statistics
correct = 0.
sum_error = 0
for i in testdata.get_targets():
if i == 1: #correct
correct += 1.
else:
sum_error += math.pow(1 - i, 2)
# percent correct
self.__pct_correct = 100 * (correct/self.__inst_test)
# root mean squared error
self.__rmse = math.sqrt(sum_error / self.__inst_test)
def testRidgeRegression(self):
cwd = os.path.dirname(os.path.abspath(sys.argv[0]))
data = LibsvmFileImporter(os.path.join(cwd,'data/regression/lin_reg'), binary=False).get_dataSet()
rr = RidgeRegression(5)
rr.trainModel(data)
#TODO: create test
self.assertTrue(True)
def testImportData(self):
cwd = os.path.dirname(os.path.abspath(sys.argv[0]))
l = LibsvmFileImporter(os.path.join(cwd,'data/classification/debug'), binary=True)
ds = l.get_dataSet()
''' contents of the debug file
-1 3:1.4324 76:1 80:1 83:1
+1 14:1 19:1.324 84:1 # A comment
# another comment
-1 73:1 75:1 76:1 80:1 85:1.155
'''
# 1. we have a result
self.assertTrue(ds is not None)
# 2. class is loaded correct
self.assertTrue(ds.get_targets(0) == -1)
# 3. comment lines correctly skipped
self.assertTrue(ds.get_targets(2) == -1)
with self.assertRaises(IndexError):
#should not exist
self.assertTrue(ds.get_targets(3) == -1)
def testImportData(self):
cwd = os.path.dirname(os.path.abspath(sys.argv[0]))
l = LibsvmFileImporter(os.path.join(cwd, 'data/classification/debug'),
binary=True)
ds = l.get_dataSet()
''' contents of the debug file
-1 3:1.4324 76:1 80:1 83:1
+1 14:1 19:1.324 84:1 # A comment
# another comment
-1 73:1 75:1 76:1 80:1 85:1.155
'''
# 1. we have a result
self.assertTrue(ds is not None)
# 2. class is loaded correct
self.assertTrue(ds.get_targets(0) == -1)
# 3. comment lines correctly skipped
self.assertTrue(ds.get_targets(2) == -1)
with self.assertRaises(IndexError):
#should not exist
self.assertTrue(ds.get_targets(3) == -1)
def testBinaryImport(self):
cwd = os.path.dirname(os.path.abspath(sys.argv[0]))
# should do
importer = LibsvmFileImporter(os.path.join(cwd,
'data/classification/a1a'),
binary=True)
importer.get_dataSet()
# should fail
with self.assertRaises(TypeError): #multi-class data
importer = LibsvmFileImporter(os.path.join(
cwd, 'data/classification/satimage.scale'),
binary=True)
sys.exit()
## process input
if classifier == None:
print "No classifier specified."
usage()
sys.exit()
if trainingFile == None:
print "No training file specified."
usage()
sys.exit()
if testFile == None:
print "No test file specified."
usage()
sys.exit()
training = LibsvmFileImporter(trainingFile, binary=True).get_dataSet()
testing = LibsvmFileImporter(testFile, binary=True).get_dataSet()
# start classification - TODO: implement report and validation
if classifier.__class__ == DualCoordinateDescent().__class__:
classifier.set_kernel(kernel)
if verbose:
print classifier
classifier.train(training.get_features(), training.get_targets())
elif classifier.__class__ == SMO_Keerthi().__class__:
classifier.set_kernel(kernel)
if verbose:
print classifier
classifier.train(training.get_features(), training.get_targets())
print "# support vectors:", classifier.get_num_support_vectors()
def buildClassifier(self, trainFile):
'''"builds" a classification model returning always 1 for each instance'''
train = LibsvmFileImporter(trainFile).get_dataSet()
self.__inst_train = train.get_numInstances()
def buildClassifier(self, trainFile):
'''"builds" a classification model returning always 1 for each instance'''
train = LibsvmFileImporter(trainFile).get_dataSet()
self.__inst_train = train.get_numInstances()
