def test_avg_auc_roc_with_splited_cv(self):
sets = split_data_cv(self.test_samples)
def tmp(train, test):
self.classifier.train(train)
return AUCROC(self.classifier, test)
aucs = [tmp(train, test) for train, test in sets]
max_auc, min_auc = (f(aucs) for f in (max, min))
avg_auc = aucroc_avg_classifier_performance(self.classifier, sets)
self.assertTrue(min_auc <= avg_auc[0] <= max_auc)
def test_split_data_cv(self):
N = 100
for _ in xrange(100):
samples = range(N)
folds = random.randint(2, N / 3)
sets = split_data_cv(samples, folds)
for train, test in sets:
for ts in test:
self.assertTrue(ts not in train)
self.assertTrue(N / folds <= len(test) <= N / folds + 1)
self.assertEqual(N, len(test) + len(train))
def test_avg_auc_roc_with_splited_cv(self):
sets = split_data_cv(self.test_samples)
def tmp(train, test):
self.classifier.train(train)
return AUCROC(self.classifier, test)
aucs = [tmp(train, test) for train, test in sets]
max_auc, min_auc = (f(aucs) for f in (max, min))
avg_auc = aucroc_avg_classifier_performance(self.classifier, sets)
self.assertTrue(min_auc <= avg_auc[0] <= max_auc)