def __init__(self, instances, feature, output_directory):
self.feature = feature
self.output_directory = path.join(output_directory, self.feature)
dir_tools.createDirectory(self.output_directory)
self.has_ground_truth = instances.hasGroundTruth()
self.generatePlotDatasets(instances)
self.setFeatureType()
def getOutputDirectories(self, al_dir, iteration_dir):
monitoring_dir = path.join(iteration_dir, 'labels_monitoring')
dir_tools.createDirectory(monitoring_dir)
evolution_dir = path.join(al_dir, 'labels_monitoring')
if self.monitoring.iteration_number == 1:
dir_tools.createDirectory(evolution_dir)
return monitoring_dir, evolution_dir
def display(self, directory):
testing_dir = path.join(directory, self.monitoring_type)
dir_tools.createDirectory(testing_dir)
self.finalComputations()
self.predictions_monitoring.display(testing_dir)
if self.has_ground_truth:
self.performance_monitoring.display(testing_dir)
if self.families_monitoring is not None:
families_dir = path.join(testing_dir, 'families')
dir_tools.createDirectory(families_dir)
self.families_monitoring.display(families_dir)
def display(self, directory):
training_dir = path.join(directory, self.monitoring_type)
dir_tools.createDirectory(training_dir)
self.finalComputations()
self.performance_monitoring.display(training_dir)
if not self.conf.families_supervision:
self.predictions_monitoring.display(training_dir)
self.coefficients.display(training_dir)
if self.families_monitoring is not None:
families_dir = path.join(training_dir, 'families')
dir_tools.createDirectory(families_dir)
self.families_monitoring.display(families_dir)
def createDataset(project, dataset):
dataset_dir = getDatasetDirectory(project, dataset)
dir_tools.createDirectory(dataset_dir)
features_dir = path.join(dataset_dir, 'features')
dir_tools.createDirectory(features_dir)
annotations_dir = path.join(dataset_dir, 'annotations')
dir_tools.createDirectory(annotations_dir)
return dataset_dir, features_dir, annotations_dir
def dumpModel(self):
# check added for Sssvdd that has no dump model function
if self.classifier.pipeline is not None:
model_dir = path.join(self.output_directory, 'model')
dir_tools.createDirectory(model_dir)
self.classifier.dumpModel(path.join(model_dir, 'model.out'))
def exportAlerts(self, datasets):
if self.classifier.alerts is None:
return
alerts_directory = path.join(self.output_directory, 'alerts')
dir_tools.createDirectory(alerts_directory)
self.classifier.alerts.export(alerts_directory)
def initComputations(self):
Iteration.initComputations(self)
dir_tools.createDirectory(self.iteration_dir)
self.monitoring.exportStartMonitoring(self.al_dir, self.iteration_dir)
def export(self):
experiment_dir = self.getOutputDirectory()
dir_tools.createDirectory(experiment_dir)
conf_filename = path.join(experiment_dir, 'conf.json')
with open(conf_filename, 'w') as f:
json.dump(self.toJson(), f, indent=2)
def setOutputDirectory(self):
self.output_directory = path.join(self.monitoring.iteration_dir,
'clustering_evaluation')
dir_tools.createDirectory(self.output_directory)
def createDatasetOutputDirectory(project, dataset):
dir_tools.createDirectory(getDatasetOutputDirectory(project, dataset))
def getEvolutionDir(self, al_dir):
evolution_dir = path.join(al_dir, 'models_performance')
if self.monitoring.iteration_number == 1:
dir_tools.createDirectory(evolution_dir)
return evolution_dir
def getMonitoringDir(self, iteration_dir):
monitoring_dir = path.join(iteration_dir, 'models_performance')
dir_tools.createDirectory(monitoring_dir)
return monitoring_dir
def createDatasetOutputDirectory(secuml_conf, project, dataset):
dir_tools.createDirectory(getDatasetOutputDirectory(secuml_conf,
project,
dataset))
