Source code for synapse.ml.featurize.CleanMissingData

# Copyright (C) Microsoft Corporation. All rights reserved.
# Licensed under the MIT License. See LICENSE in project root for information.


import sys
if sys.version >= '3':
    basestring = str

from pyspark import SparkContext, SQLContext
from pyspark.sql import DataFrame
from pyspark.ml.param.shared import *
from pyspark import keyword_only
from pyspark.ml.util import JavaMLReadable, JavaMLWritable
from synapse.ml.core.platform import running_on_synapse_internal
from synapse.ml.core.serialize.java_params_patch import *
from pyspark.ml.wrapper import JavaTransformer, JavaEstimator, JavaModel
from pyspark.ml.evaluation import JavaEvaluator
from pyspark.ml.common import inherit_doc
from synapse.ml.core.schema.Utils import *
from pyspark.ml.param import TypeConverters
from synapse.ml.core.schema.TypeConversionUtils import generateTypeConverter, complexTypeConverter
from synapse.ml.featurize.CleanMissingDataModel import CleanMissingDataModel

[docs]@inherit_doc class CleanMissingData(ComplexParamsMixin, JavaMLReadable, JavaMLWritable, JavaEstimator): """ Args: cleaningMode (str): Cleaning mode customValue (str): Custom value for replacement inputCols (list): The names of the input columns outputCols (list): The names of the output columns """ cleaningMode = Param(Params._dummy(), "cleaningMode", "Cleaning mode", typeConverter=TypeConverters.toString) customValue = Param(Params._dummy(), "customValue", "Custom value for replacement", typeConverter=TypeConverters.toString) inputCols = Param(Params._dummy(), "inputCols", "The names of the input columns", typeConverter=TypeConverters.toListString) outputCols = Param(Params._dummy(), "outputCols", "The names of the output columns", typeConverter=TypeConverters.toListString) @keyword_only def __init__( self, java_obj=None, cleaningMode="Mean", customValue=None, inputCols=None, outputCols=None ): super(CleanMissingData, self).__init__() if java_obj is None: self._java_obj = self._new_java_obj("com.microsoft.azure.synapse.ml.featurize.CleanMissingData", self.uid) else: self._java_obj = java_obj self._setDefault(cleaningMode="Mean") if hasattr(self, "_input_kwargs"): kwargs = self._input_kwargs else: kwargs = self.__init__._input_kwargs if java_obj is None: for k,v in kwargs.items(): if v is not None: getattr(self, "set" + k[0].upper() + k[1:])(v)
[docs] @keyword_only def setParams( self, cleaningMode="Mean", customValue=None, inputCols=None, outputCols=None ): """ Set the (keyword only) parameters """ if hasattr(self, "_input_kwargs"): kwargs = self._input_kwargs else: kwargs = self.__init__._input_kwargs return self._set(**kwargs)
[docs] @classmethod def read(cls): """ Returns an MLReader instance for this class. """ return JavaMMLReader(cls)
[docs] @staticmethod def getJavaPackage(): """ Returns package name String. """ return "com.microsoft.azure.synapse.ml.featurize.CleanMissingData"
@staticmethod def _from_java(java_stage): module_name=CleanMissingData.__module__ module_name=module_name.rsplit(".", 1)[0] + ".CleanMissingData" return from_java(java_stage, module_name)
[docs] def setCleaningMode(self, value): """ Args: cleaningMode: Cleaning mode """ self._set(cleaningMode=value) return self
[docs] def setCustomValue(self, value): """ Args: customValue: Custom value for replacement """ self._set(customValue=value) return self
[docs] def setInputCols(self, value): """ Args: inputCols: The names of the input columns """ self._set(inputCols=value) return self
[docs] def setOutputCols(self, value): """ Args: outputCols: The names of the output columns """ self._set(outputCols=value) return self
[docs] def getCleaningMode(self): """ Returns: cleaningMode: Cleaning mode """ return self.getOrDefault(self.cleaningMode)
[docs] def getCustomValue(self): """ Returns: customValue: Custom value for replacement """ return self.getOrDefault(self.customValue)
[docs] def getInputCols(self): """ Returns: inputCols: The names of the input columns """ return self.getOrDefault(self.inputCols)
[docs] def getOutputCols(self): """ Returns: outputCols: The names of the output columns """ return self.getOrDefault(self.outputCols)
def _create_model(self, java_model): try: model = CleanMissingDataModel(java_obj=java_model) model._transfer_params_from_java() except TypeError: model = CleanMissingDataModel._from_java(java_model) return model def _fit(self, dataset): java_model = self._fit_java(dataset) return self._create_model(java_model)