def run(fast=False):
# read data
train_path = os.path.join(HOME_DIR, 'resources', 'datasets', 'titanic-train.csv')
df_train = pandas.read_csv(train_path)
num, cat = FeatureTypes.numerical, FeatureTypes.categorical
dmd_train, dmd_test = DMD.from_df(df_train=df_train, df_test=None,
is_classification=True,
target_name='Survived',
feature_types=[num, cat, cat, cat, num, num, num, cat, num, cat, cat],
categorical_encoding=True, nan_list=['?'],
split_ratio=0.2)
classifier = Pipeline(steps=[('Imputer', SimpleImputer()),
('Estimator', RandomForestClassifier(n_estimators=3))])
classifier.fit(dmd_train.values, dmd_train.target)
pytrust = PyTrust(
model=classifier,
xtrain=dmd_train,
xtest=dmd_test,
metric='recall')
# some analysis
print('\n'.join(pytrust.insights))
pytrust.dataset_analysis_report.plot()
pytrust.scoring_report.plot()
pytrust.sensitivity_report.plot()
def test_dummy(self):
dmd_train = DMD(x=numpy.zeros((3, 3)), y=numpy.zeros((3, 1)))
dmd_test = DMD(x=numpy.zeros((3, 3)), y=numpy.zeros((3, 1)))
print('init model')
model = DMDPipeline(transformers=[TransformerDummy],
estimator=EstimatorDummy,
hyperparameters=[{
'name': 'Dummy'
}, {
'n_estimators': 10
}])
self.assertIsInstance(model.name, str)
self.assertIsInstance(model.transformers, list)
self.assertEqual(len(model.transformers), 1)
self.assertIsInstance(model.transformers[0], TransformerBase)
self.assertEqual(model.transformers[0].name, 'Dummy')
self.assertIsInstance(model.estimator, EstimatorBase)
self.assertEqual(model.estimator.kwargs['n_estimators'], 10)
self.assertEqual(len(model.hyperparameters), 2)
self.assertEqual((model.transformers[0].name, {
'name': 'Dummy'
}), model.hyperparameters[0])
self.assertEqual((model.estimator.name, {
'n_estimators': 10
}), model.hyperparameters[1])
print('fit model')
model.fit(dmd_train)
print('model predict')
predictions = model.predict(dmd_test)
self.assertIsInstance(predictions, DMD)
def predict_proba(self, dmd: DMD, **kwargs):
print('EstimatorDummy', 'predict_proba', 'n_estimators',
self.kwargs.get('n_estimators'))
return DMD(x=dmd.values[:, 0:2])
def fit(self, dmd: DMD, **kwargs):
print('EstimatorDummy', 'fit', 'n_estimators',
self.kwargs.get('n_estimators'))
pass
def predict(self, dmd: DMD, **kwargs):
print('EstimatorDummy', 'predict', 'n_estimators',
self.kwargs.get('n_estimators'))
return DMD(x=dmd.values[:, 0:1])
def predict_proba(self, dmd: DMD, **kwargs):
print('EstimatorDummy', 'predict_proba', 'n_estimators',
self.kwargs.get('n_estimators'))
return DMD(x=dmd.values[:, 0:2])
dmd_train = DMD(x=numpy.zeros((3, 3)), y=numpy.zeros((3, 1)))
dmd_test = DMD(x=numpy.zeros((3, 3)), y=numpy.zeros((3, 1)))
print('init model')
model = DMDPipeline(transformers=[TransformerDummy],
estimator=EstimatorDummy,
hyperparameters=[{
'name': 'Dummy'
}, {
'n_estimators': 10
}])
print('fit model')
model.fit(dmd_train)
print('model predict')