rxSerializeModel: RevoScaleR Model Serialization and Unserialization


Serialize a RevoScaleR/MicrosoftML model in raw format to enable saving the model. This allows model to be loaded into SQL Server for real-time scoring.


  rxSerializeModel(model, metadata = NULL, realtimeScoringOnly = FALSE, ...)

  rxUnserializeModel(serializedModel, ...)



RevoScaleR/MicrosoftML model to be serialized


Arbitrary metadata of raw type to be stored with the serialized model. Metadata will be returned when unserialized.


Drops fields not required for real-time scoring. NOTE: Setting this flag could reduce the model size but rxUnserializeModel can no longer retrieve the RevoScaleR model


Serialized model to be unserialized


rxSerializeModel converts models into raw bytes to allow them to be saved and used for real-time scoring.

The following is the list of models that are currently supported in real-time scoring:

* RevoScaleR

* rxLogit

* rxLinMod

* rxBTrees

* rxDTree

* rxDForest

* MicrosoftML

* rxFastTrees

* rxFastForest

* rxLogisticRegression

* rxOneClassSvm

* rxNeuralNet

* rxFastLinear

RevoScaleR models containing R transformations or transform based formula (e.g "A ~ log(B)") are not supported in real-time scoring. Such transforms to input data may need to be handled before calling real-time scoring.

rxUnserializeModel method is used to retrieve the original R model object and metadata from the serialized raw model.


rxSerializeModel returns a serialized model.

rxUnserializeModel returns original R model object. If metadata is also present returns a list containing the original model object and metadata.


Microsoft Corporation Microsoft Technical Support

See Also


 myIris <- iris
 form <- Sepal.Length ~ Sepal.Width + Petal.Length + Petal.Width + Species
 irisLinMod <- rxLinMod(form, data = myIris)

 # Serialize model for scoring
 serializedModel <- rxSerializeModel(irisLinMod)

 # Save model to file or SQL Server (use rxWriteObject)
 # serialized model can now be used for real-time scoring

 unserializedModel <- rxUnserializeModel(serializedModel)