def test_invalids(self):
estimator, pd, pt = _set_parameters_and_estimator(
'RandomForestClassifier', self.table_chard_fp, self.md_chard_fp,
'Region', n_estimators=10, n_jobs=1, cv=1,
random_state=123, parameter_tuning=False, classification=True)
regressor, pd, pt = _set_parameters_and_estimator(
'RandomForestRegressor', self.table_chard_fp, self.md_chard_fp,
'Region', n_estimators=10, n_jobs=1, cv=1,
random_state=123, parameter_tuning=False, classification=True)
# zero samples (if mapping file and table have no common samples)
with self.assertRaisesRegex(ValueError, "metadata"):
estimator, cm, accuracy, importances = split_optimize_classify(
self.table_ecam_fp, self.md_chard_fp, 'Region', estimator,
self.temp_dir.name, test_size=0.5, cv=1, random_state=123,
n_jobs=1, optimize_feature_selection=False,
parameter_tuning=False, param_dist=None,
calc_feature_importance=False)
# too few samples to stratify
with self.assertRaisesRegex(ValueError, "metadata"):
estimator, cm, accuracy, importances = split_optimize_classify(
self.table_chard_fp, self.md_chard_fp, 'Region', estimator,
self.temp_dir.name, test_size=0.9, cv=1, random_state=123,
n_jobs=1, optimize_feature_selection=False,
parameter_tuning=False, param_dist=None,
calc_feature_importance=False)
# regressor chosen for classification problem
with self.assertRaisesRegex(ValueError, "convert"):
estimator, cm, accuracy, importances = split_optimize_classify(
self.table_chard_fp, self.md_chard_fp, 'Region', regressor,
self.temp_dir.name, test_size=0.5, cv=1, random_state=123,
n_jobs=1, optimize_feature_selection=False,
parameter_tuning=False, param_dist=None,
calc_feature_importance=False)
def test_invalids(self):
estimator, pad, pt = _set_parameters_and_estimator(
'RandomForestClassifier',
self.table_chard_fp,
self.md_chard_fp,
'Region',
n_estimators=10,
n_jobs=1,
cv=1,
random_state=123,
parameter_tuning=False,
classification=True,
missing_samples='ignore')
regressor, pad, pt = _set_parameters_and_estimator(
'RandomForestRegressor',
self.table_chard_fp,
self.md_chard_fp,
'Region',
n_estimators=10,
n_jobs=1,
cv=1,
random_state=123,
parameter_tuning=False,
classification=True,
missing_samples='ignore')
def test_feature_ordering(self):
# replicate minimal split_optimize_classify to extract importances
estimator, pad, pt = _set_parameters_and_estimator(
'RandomForestRegressor',
self.table_ecam_fp,
self.md_ecam_fp,
'month',
n_estimators=10,
n_jobs=1,
cv=1,
random_state=123,
parameter_tuning=False,
classification=False)
X_train, X_test, y_train, y_test = _prepare_training_data(
self.table_ecam_fp,
self.md_ecam_fp,
'month',
test_size=0.1,
random_state=123,
load_data=True,
stratify=False)
X_train, X_test, importance = _optimize_feature_selection(
self.temp_dir.name,
X_train,
X_test,
y_train,
estimator,
cv=3,
step=0.2,
n_jobs=1)
estimator, accuracy, y_pred = _fit_and_predict(
X_train,
X_test,
y_train,
y_test,
estimator,
scoring=mean_squared_error)
# pull important features from a different dataframe
importances = _calculate_feature_importances(X_train, estimator)
table = self.table_ecam_fp.loc[:, importances["feature"]]
# confirm ordering of feature (column) names
ca = list(X_train.columns.values)
cb = list(table.columns.values)
self.assertEqual(ca, cb)
def test_regressors(self):
for regressor in [
'RandomForestRegressor', 'ExtraTreesRegressor',
'GradientBoostingRegressor', 'AdaBoostRegressor', 'Lasso',
'Ridge', 'ElasticNet', 'KNeighborsRegressor', 'LinearSVR',
'SVR'
]:
tmpd = join(self.temp_dir.name, regressor)
mkdir(tmpd)
estimator, pad, pt = _set_parameters_and_estimator(
regressor,
self.table_ecam_fp,
self.md_ecam_fp,
'month',
n_estimators=10,
n_jobs=1,
cv=1,
random_state=123,
parameter_tuning=False,
classification=False)
estimator, cm, accuracy, importances = split_optimize_classify(
self.table_ecam_fp,
self.md_ecam_fp,
'month',
estimator,
tmpd,
test_size=0.5,
cv=1,
random_state=123,
n_jobs=1,
optimize_feature_selection=False,
parameter_tuning=False,
param_dist=None,
classification=False,
calc_feature_importance=False,
scoring=mean_squared_error)
self.assertAlmostEqual(
accuracy,
seeded_results[regressor],
places=4,
msg='Accuracy of %s regressor was %f, but expected %f' %
(regressor, accuracy, seeded_results[regressor]))
def test_classifiers(self):
for classifier in ['RandomForestClassifier', 'ExtraTreesClassifier',
'GradientBoostingClassifier', 'AdaBoostClassifier',
'LinearSVC', 'SVC', 'KNeighborsClassifier']:
tmpd = join(self.temp_dir.name, classifier)
mkdir(tmpd)
estimator, pd, pt = _set_parameters_and_estimator(
classifier, self.table_chard_fp, self.md_chard_fp, 'Region',
n_estimators=10, n_jobs=1, cv=1,
random_state=123, parameter_tuning=False, classification=True)
estimator, cm, accuracy, importances = split_optimize_classify(
self.table_chard_fp, self.md_chard_fp, 'Region', estimator,
tmpd, test_size=0.5, cv=1, random_state=123,
n_jobs=1, optimize_feature_selection=False,
parameter_tuning=False, param_dist=None,
calc_feature_importance=False)
self.assertAlmostEqual(accuracy, seeded_results[classifier])
self.assertAlmostEqual(
accuracy, seeded_results[classifier], places=4,
msg='Accuracy of %s classifier was %f, but expected %f' % (
classifier, accuracy, seeded_results[classifier]))
