def __init__(self, X, Y, parameters, conveyor):
# Initialize parent class
try:
BaseEstimator.__init__(self, X, Y, parameters, conveyor)
LOG.debug('Initialize BaseEstimator parent class')
except Exception as e:
LOG.error(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
self.conveyor.setError(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
return
# Load estimator parameters
self.estimator_parameters = self.param.getDict('PLSR_parameters')
# Scale is hard-coded to False for making use of external scalers
self.estimator_parameters['scale'] = False
self.name = "PLSR"
# Check if the model is quantitative
if not self.param.getVal('quantitative'):
LOG.error('PLSR only applies to quantitative data')
self.conveyor.setError('PLSR only applies to quantitative data')
return
def __init__(self, X, Y, parameters, conveyor):
# Initialize parent class
try:
BaseEstimator.__init__(self, X, Y, parameters, conveyor)
LOG.debug('Initialize BaseEstimator parent class')
except Exception as e:
LOG.error(f'Error initializing BaseEstimator parent class with exception: {e}')
self.conveyor.setError(f'Error initializing BaseEstimator parent class with exception: {e}')
return
# Load estimator parameters
self.estimator_parameters = self.param.getDict('SVM_parameters')
# Load tune parameters
self.tune_parameters = self.param.getDict('SVM_optimize')
if self.param.getVal('quantitative'):
# Remove parameters of SVC class and set the name
self.name = "SVM-R"
self.estimator_parameters.pop("class_weight", None)
self.estimator_parameters.pop("probability", None)
self.estimator_parameters.pop("decision_function_shape", None)
self.estimator_parameters.pop("random_state", None)
self.tune_parameters.pop("class_weight", None)
self.tune_parameters.pop("random_state", None)
self.tune_parameters.pop("probability", None)
else:
# Remove parameters of SVR class and set the name
self.estimator_parameters.pop("epsilon", None)
self.name = "SVM-C"
def __init__(self, X, Y, parameters, conveyor):
# Initialize parent class
try:
BaseEstimator.__init__(self, X, Y, parameters, conveyor)
LOG.debug('Initialize BaseEstimator parent class')
except Exception as e:
self.conveyor.setError(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
LOG.error(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
return
# Load estimator parameters
self.estimator_parameters = self.param.getDict('XGBOOST_parameters')
# Load tune parameters
self.tune_parameters = self.param.getDict('XGBOOST_optimize')
if self.param.getVal('quantitative'):
self.estimator_parameters['objective'] = 'reg:squarederror'
self.name = "XGB-Regressor"
else:
self.estimator_parameters['objective'] = 'binary:logistic'
self.name = "XGB-Classifier"
# Missing value must be defined. Otherwyse it returns 'nan' which cannot be
# converted to JSON and produces trouble in different points
self.estimator_parameters['missing'] = -99.99999
def __init__(self, X, Y, parameters, conveyor):
# Initialize parent class
try:
BaseEstimator.__init__(self, X, Y, parameters, conveyor)
LOG.debug('Initialize BaseEstimator parent class')
except Exception as e:
self.conveyor.setError(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
LOG.error(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
return
# Load estimator parameters
self.estimator_parameters = self.param.getDict('RF_parameters')
# Load tune parameters
self.tune_parameters = self.param.getDict('RF_optimize')
if self.param.getVal('quantitative'):
self.name = "RF-R"
self.tune_parameters.pop("class_weight")
self.estimator_parameters.pop("class_weight")
else:
self.name = "RF-C"
def __init__(self, X, Y, parameters, conveyor):
# Initialize parent class
try:
BaseEstimator.__init__(self,X, Y, parameters, conveyor)
LOG.debug('Initialize BaseEstimator parent class')
except Exception as e:
LOG.error(f'Error initializing BaseEstimator parent class with exception: {e}')
self.conveyor.setError(f'Error initializing BaseEstimator parent class with exception: {e}')
return
# Load estimator parameters
self.estimator_parameters = self.param.getDict('PLSDA_parameters')
# Solves back-comptibility issue
if 'optimize' in self.estimator_parameters:
self.estimator_parameters.pop("optimize")
# Scale is hard-coded to False for making use of external scalers
self.estimator_parameters['scale'] = False
self.name = "PLSDA"
if 'threshold' in self.estimator_parameters:
self.threshold = self.estimator_parameters['threshold']
else:
self.threshold = 0.5
if self.param.getVal('quantitative'):
self.conveyor.setError('PLSDA only applies to qualitative data')
return
# For confidential models, create an empty estimator
if self.param.getVal('confidential'):
self.estimator = PLS_da(**self.estimator_parameters)
def __init__(self, X, Y, parameters, conveyor):
# Initialize parent class
try:
BaseEstimator.__init__(self, X, Y, parameters, conveyor)
LOG.debug('Initialize BaseEstimator parent class')
except Exception as e:
LOG.error(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
self.conveyor.setError(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
return
# Load estimator parameters
self.estimator_parameters = self.param.getDict('PLSDA_parameters')
# Scale is hard-coded to False for making use of external scalers
self.estimator_parameters['scale'] = False
self.name = "PLSDA"
if self.param.getVal('quantitative'):
LOG.error('PLSDA only applies to qualitative data')
self.conveyor.setError('PLSDA only applies to qualitative data')
return
if self.param.getVal('conformal'):
LOG.error('Conformal prediction no implemented in PLSDA yet')
self.conveyor.setError(
'Conformal prediction no implemented in PLSDA yet')
return
def __init__(self, X, Y, parameters, conveyor):
# Initialize parent class
try:
BaseEstimator.__init__(self, X, Y, parameters, conveyor)
LOG.debug('Initialize BaseEstimator parent class')
except Exception as e:
LOG.error(f'Error initializing BaseEstimator parent'
f'class with exception: {e}')
raise e
self.method_name = ''
def __init__(self, X, Y, parameters, conveyor):
# Initialize parent class
try:
BaseEstimator.__init__(self, X, Y, parameters, conveyor)
LOG.debug('Initializing BaseEstimator parent class')
except Exception as e:
self.conveyor.setError(f'Error initializing BaseEstimator parent class with exception: {e}')
LOG.error(f'Error initializing BaseEstimator parent class with exception: {e}')
return
# Load estimator parameters
self.estimator_parameters = self.param.getDict('GNB_parameters')
if self.param.getVal('quantitative'):
self.conveyor.setError('GNB only applies to qualitative data')
LOG.error('GNB only applies to qualitative data')
else:
self.name = "GNB-Classifier"
def __init__(self, X, Y, parameters, conveyor):
# Initialize parent class
try:
BaseEstimator.__init__(self, X, Y, parameters, conveyor)
LOG.debug('Initializing BaseEstimator parent class')
except Exception as e:
self.conveyor.setError(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
LOG.error(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
return
# Load estimator parameters
GNB_parameters = self.param.getDict('GNB_parameters')
priors = [0.0, 0.0]
if 'prior_negative' in GNB_parameters and GNB_parameters[
'prior_negative'] != None:
priors[0] = GNB_parameters['prior_negative']
if 'prior_positive' in GNB_parameters and GNB_parameters[
'prior_positive'] != None:
priors[1] = GNB_parameters['prior_positive']
self.estimator_parameters = {}
if GNB_parameters['var_smoothing'] is not None:
self.estimator_parameters['var_smoothing'] = GNB_parameters[
'var_smoothing']
if priors[0] != 0.0 and priors[1] != 0.0:
if priors[0] + priors[1] != 1.0:
LOG.error(
f'GNB: the sum of the priors should be 1. priors set to {priors[0], priors[1]} '
)
priors[1] = 1.0 - priors[0]
self.estimator_parameters['priors'] = priors
# self.param.setInnerVal('GNB_parameters','priors',priors)
if self.param.getVal('quantitative'):
self.conveyor.setError('GNB only applies to qualitative data')
LOG.error('GNB only applies to qualitative data')
else:
self.name = "GNB-Classifier"
def __init__(self, X, Y, parameters, conveyor):
# Initialize parent class
try:
BaseEstimator.__init__(self, X, Y, parameters, conveyor)
LOG.debug('Initialize BaseEstimator parent class')
except Exception as e:
self.conveyor.setError(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
LOG.error(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
return
# Load estimator parameters
self.estimator_parameters = self.param.getDict('RF_parameters')
# patch to solve bug in the object_type definition of max_depth
# should not appear from versions older than 20/01/2021
if 'max_depth' in self.estimator_parameters:
v = self.estimator_parameters['max_depth']
if v is not None:
try:
self.estimator_parameters['max_depth'] = int(v)
except:
self.estimator_parameters['max_depth'] = None
# Load tune parameters
self.tune_parameters = self.param.getDict('RF_optimize')
if self.param.getVal('quantitative'):
self.name = "RF-R"
self.tune_parameters.pop("class_weight")
self.estimator_parameters.pop("class_weight")
else:
self.name = "RF-C"
def __init__(self, X, Y, parameters, conveyor):
# Initialize parent class
try:
BaseEstimator.__init__(self, X, Y, parameters, conveyor)
LOG.debug('Initialize BaseEstimator parent class')
except Exception as e:
self.conveyor.setError(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
LOG.error(
f'Error initializing BaseEstimator parent class with exception: {e}'
)
return
# Load estimator parameters
self.estimator_parameters = self.param.getDict('Keras_parameters')
# Load tune parameters
self.tune_parameters = self.param.getDict('Keras_optimize')
if self.param.getVal('quantitative'):
self.name = "Keras-Regressor"
else:
self.name = "Keras-Classifier"