def test_linear_regression(self):
trained_linear_model = self.regression_trainer.linear_regression()
self.assertIsInstance(trained_linear_model, TrainedSupervisedModel)
result = trained_linear_model.metrics
helpers.assertBetween(self, 500, 750, result['mean_squared_error'])
helpers.assertBetween(self, 18, 29, result['mean_absolute_error'])
def test_knn(self):
trained_knn = self.classification_trainer.knn()
result = trained_knn.metrics
self.assertIsInstance(trained_knn, TrainedSupervisedModel)
helpers.assertBetween(self, 0.5, 0.95, result['roc_auc'])
helpers.assertBetween(self, 0.79, 0.95, result['accuracy'])
def test_random_forest_classification(self):
# Force plot to save to prevent blocking
trained_random_forest = self.classification_trainer.random_forest_classification(save_plot=True)
result = trained_random_forest.metrics
self.assertIsInstance(trained_random_forest, TrainedSupervisedModel)
helpers.assertBetween(self, 0.65, 0.95, result['roc_auc'])
helpers.assertBetween(self, 0.8, 0.95, result['accuracy'])
def test_logistic_regression(self):
trained_lr = self.classification_trainer.logistic_regression()
self.assertIsInstance(trained_lr, TrainedSupervisedModel)
result = trained_lr.metrics
helpers.assertBetween(self, 0.52, 0.95, result['roc_auc'])
helpers.assertBetween(self, 0.6, 0.95, result['accuracy'])
def test_logistic_regression(self):
trained_lr = self.classification_trainer.logistic_regression()
self.assertIsInstance(trained_lr, TrainedSupervisedModel)
result = trained_lr.metrics
helpers.assertBetween(self, 0.52, 0.95, result['roc_auc'])
helpers.assertBetween(self, 0.6, 0.95, result['accuracy'])
def test_ensemble_classification(self):
trained_ensemble = self.classification_trainer.ensemble()
self.assertIsInstance(trained_ensemble, TrainedSupervisedModel)
result = trained_ensemble.metrics
helpers.assertBetween(self, 0.6, 0.97, result['roc_auc'])
helpers.assertBetween(self, 0.6, 0.97, result['accuracy'])
def test_random_forest_regression(self):
trained_rf_regressor = self.regression_trainer.random_forest_regression()
self.assertIsInstance(trained_rf_regressor, TrainedSupervisedModel)
result = trained_rf_regressor.metrics
helpers.assertBetween(self, 350, 750, result['mean_squared_error'])
helpers.assertBetween(self, 10, 25, result['mean_absolute_error'])
def test_linear_regression(self):
trained_linear_model = self.regression_trainer.linear_regression()
self.assertIsInstance(trained_linear_model, TrainedSupervisedModel)
result = trained_linear_model.metrics
helpers.assertBetween(self, 450, 800, result['mean_squared_error'])
helpers.assertBetween(self, 16, 29, result['mean_absolute_error'])
def test_ensemble_classification(self):
trained_ensemble = self.classification_trainer.ensemble()
self.assertIsInstance(trained_ensemble, TrainedSupervisedModel)
result = trained_ensemble.metrics
helpers.assertBetween(self, 0.6, 0.95, result['roc_auc'])
helpers.assertBetween(self, 0.6, 0.95, result['accuracy'])
def test_random_forest_regression(self):
trained_rf_regressor = self.regression_trainer.random_forest_regression()
self.assertIsInstance(trained_rf_regressor, TrainedSupervisedModel)
result = trained_rf_regressor.metrics
helpers.assertBetween(self, 400, 700, result['mean_squared_error'])
helpers.assertBetween(self, 10, 20, result['mean_absolute_error'])
def test_knn(self):
trained_knn = self.classification_trainer.knn()
result = trained_knn.metrics
self.assertIsInstance(trained_knn, TrainedSupervisedModel)
helpers.assertBetween(self, 0.5, 0.85, result['roc_auc'])
helpers.assertBetween(self, 0.79, 0.95, result['accuracy'])
def test_pr(self):
df = pd.DataFrame({'a': np.repeat(np.arange(.1, 1.1, .1), 10)})
b = np.repeat(0, 100)
b[[56, 62, 63, 68, 74, 75, 76, 81, 82, 84, 85, 87, 88] + list(range(90, 100))] = 1
df['b'] = b
# PR_AUC
out = hcai_eval.compute_pr(df['b'], df['a'])
test_helpers.assertBetween(self, 0.8, 0.87, out['pr_auc'])
self.assertAlmostEqual(round(out['best_precision'], 4), 0.8000)
self.assertAlmostEqual(round(out['best_recall'], 4), 0.6957)
def test_pr(self):
df = pd.DataFrame({'a': np.repeat(np.arange(.1, 1.1, .1), 10)})
b = np.repeat(0, 100)
b[[56, 62, 63, 68, 74, 75, 76, 81, 82, 84, 85, 87, 88] + list(range(90, 100))] = 1
df['b'] = b
# PR_AUC
out = hcai_eval.compute_pr(df['b'], df['a'])
test_helpers.assertBetween(self, 0.8, 0.87, out['pr_auc'])
self.assertAlmostEqual(round(out['best_precision'], 4), 0.8000)
self.assertAlmostEqual(round(out['best_recall'], 4), 0.6957)
def test_random_forest_classification(self):
# Force plot to save to prevent matplotlib blocking during testing
trained_random_forest = self.classification_trainer.random_forest_classification(save_plot=True)
result = trained_random_forest.metrics
self.assertIsInstance(trained_random_forest, TrainedSupervisedModel)
helpers.assertBetween(self, 0.65, 0.95, result['roc_auc'])
helpers.assertBetween(self, 0.8, 0.95, result['accuracy'])
# Clean up saved plot (see note above)
try:
os.remove('FeatureImportances.png')
except OSError:
pass
def test_random_forest_classification(self):
# Force plot to save to prevent blocking when testing
trained_random_forest = self.classification_trainer.random_forest_classification(save_plot=True)
result = trained_random_forest.metrics
self.assertIsInstance(trained_random_forest, TrainedSupervisedModel)
helpers.assertBetween(self, 0.65, 0.95, result['roc_auc'])
helpers.assertBetween(self, 0.8, 0.95, result['accuracy'])
# Clean up saved plot (see note above)
try:
os.remove('FeatureImportances.png')
except OSError:
pass
def test_logistic_regression_no_tuning(self):
self.assertIsInstance(self.lr, TrainedSupervisedModel)
test_helpers.assertBetween(self, 0.5, 0.8, self.lr.metrics['roc_auc'])
def test_random_forest_tuning(self):
rf = self.trainer.random_forest_classifier(randomized_search=True)
self.assertIsInstance(rf, TrainedSupervisedModel)
test_helpers.assertBetween(self, 0.7, 0.97, rf.metrics['roc_auc'])
def test_logistic_regression_no_tuning(self):
self.assertIsInstance(self.lr, TrainedSupervisedModel)
test_helpers.assertBetween(self, 0.5, 0.8, self.lr.metrics['roc_auc'])
def test_random_forest_tuning(self):
rf = self.trainer.random_forest_classifier(randomized_search=True)
self.assertIsInstance(rf, TrainedSupervisedModel)
test_helpers.assertBetween(self, 0.7, 0.97, rf.metrics['roc_auc'])
