rxEnsemble: Ensembles

Train an ensemble of models

Usage

  rxEnsemble(formula = NULL, data, trainers, type = c("binary", "regression",
    "multiClass", "anomaly"), randomSeed = NULL,
    modelCount = length(trainers), replace = FALSE, sampRate = NULL,
    splitData = FALSE, combineMethod = c("median", "average", "vote"),
    maxCalibration = 1e+05, mlTransforms = NULL, mlTransformVars = NULL,
    rowSelection = NULL, transforms = NULL, transformObjects = NULL,
    transformFunc = NULL, transformVars = NULL, transformPackages = NULL,
    transformEnvir = NULL, blocksPerRead = rxGetOption("blocksPerRead"),
    reportProgress = rxGetOption("reportProgress"), verbose = 1,
    computeContext = rxGetOption("computeContext"), ...)

Arguments

formula

The formula as described in rxFormula. Interaction terms and F() are not currently supported in the MicrosoftML.

data

A data source object or a character string specifying a .xdffile or a data frame object. Alternatively, it can be a list of data sources indicating each model should be trained using one of the data sources in the list. In this case, the length of the data list must be equal to modelCount.

trainers

A list of trainers with their arguments. The trainers are created by using fastTrees, fastForest, fastLinear, logisticRegression or neuralNet.

type

A character string that specifies the type of ensemble: "binary" for Binary Classification or "regression" for Regression.

randomSeed

Specifies the random seed. The default value is NULL.

modelCount

Specifies the number of models to train. If this number is greater than the length of the trainers list, the trainers list is duplicated to match modelCount.

replace

A logical value specifying if the sampling of observations should be done with or without replacement. The default value is /codeFALSE.

sampRate

a scalar of positive value specifying the percentage of observations to sample for each trainer. The default is 1.0 for sampling with replacement (i.e., replace=TRUE) and 0.632 for sampling without replacement (i.e., replace=FALSE). When splitData is TRUE, the default of sampRate is 1.0 (no sampling is done before splitting).

splitData

A logical value specifying whether or not to train the base models on non-overlapping partitions. The default is FALSE. It is available only for RxSpark compute context and ignored for others.

combineMethod

Specifies the method used to combine the models:

  • median to compute the median of the individual model outputs,
  • average to compute the average of the individual model outputs and
  • vote to compute (pos-neg) / the total number of models, where 'pos' is the number of positive outputs and 'neg' is the number of negative outputs.

maxCalibration

Specifies the maximum number of examples to use for calibration. This argument is ignored for all tasks other than binary classification.

mlTransforms

Specifies a list of MicrosoftML transforms to be performed on the data before training or NULL if no transforms are to be performed. Transforms that require an additional pass over the data (such as featurizeText, categorical) are not allowed. These transformations are performed after any specified R transformations. The default value is NULL.

mlTransformVars

Specifies a character vector of variable names to be used in mlTransforms or NULL if none are to be used. The default value is NULL.

rowSelection

Specifies the rows (observations) from the data set that are to be used by the model with the name of a logical variable from the data set (in quotes) or with a logical expression using variables in the data set. For example, rowSelection = "old" will only use observations in which the value of the variable old is TRUE. rowSelection = (age > 20) & (age < 65) & (log(income) > 10) only uses observations in which the value of the age variable is between 20 and 65 and the value of the log of the income variable is greater than 10. The row selection is performed after processing any data transformations (see the arguments transforms or transformFunc). As with all expressions, rowSelection can be defined outside of the function call using the expression function.

transforms

An expression of the form list(name = expression, ``...) that represents the first round of variable transformations. As with all expressions, transforms (or rowSelection) can be defined outside of the function call using the expression function. The default value is NULL.

transformObjects

A named list that contains objects that can be referenced by transforms, transformsFunc, and rowSelection. The default value is NULL.

transformFunc

The variable transformation function. See rxTransform for details. The default value is NULL.

transformVars

A character vector of input data set variables needed for the transformation function. See rxTransform for details. The default value is NULL.

transformPackages

A character vector specifying additional R packages (outside of those specified in rxGetOption("transformPackages")) to be made available and preloaded for use in variable transformation functions. For example, those explicitly defined in RevoScaleR functions via their transforms and transformFunc arguments or those defined implicitly via their formula or rowSelection arguments. The transformPackages argument may also be NULL, indicating that no packages outside rxGetOption("transformPackages") are preloaded. The default value is NULL.

transformEnvir

A user-defined environment to serve as a parent to all environments developed internally and used for variable data transformation. If transformEnvir = NULL, a new "hash" environment with parent baseenv() is used instead. The default value is NULL.

blocksPerRead

Specifies the number of blocks to read for each chunk of data read from the data source.

reportProgress

An integer value that specifies the level of reporting on the row processing progress:

  • 0: no progress is reported.
  • 1: the number of processed rows is printed and updated.
  • 2: rows processed and timings are reported.
  • 3: rows processed and all timings are reported.

verbose

An integer value that specifies the amount of output wanted. If 0, no verbose output is printed during calculations. Integer values from 1 to 4 provide increasing amounts of information. The default value is 1.

computeContext

Sets the context in which computations are executed, specified with a valid RxComputeContext. Currently local and RxSpark compute contexts are supported. When RxSpark is specified, the training of the models is done in a distributed way, and the ensembling is done locally. Note that the compute context cannot be non-waiting.

...

Additional arguments to be passed directly to the Microsoft Compute Engine.

Details

/coderxEnsemble is a function that trains a number of models of various kinds to obtain better predictive performance than could be obtained from a single model.

Value

A rxEnsemble object with the trained ensemble model.

Examples


 # Create an ensemble of regression rxFastTrees models

 # use xdf data source
 dataFile <- file.path(rxGetOption("sampleDataDir"), "claims4blocks.xdf")
 rxGetInfo(dataFile, getVarInfo = TRUE, getBlockSizes = TRUE)
 form <- cost ~ age + type + number

 rxSetComputeContext("localpar")
 rxGetComputeContext()

 # build an ensemble model that contains three 'rxFastTrees' models with different parameters
 ensemble <- rxEnsemble(
     formula = form,
     data = dataFile,
     type = "regression",
     trainers = list(fastTrees(), fastTrees(numTrees = 60), fastTrees(learningRate = 0.1)), #a list of trainers with their arguments.
     replace = TRUE # Indicates using a bootstrap sample for each trainer
     )

 # use text data source
 colInfo <- list(DayOfWeek = list(type = "factor", levels = c("Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday")))

 source <- system.file("SampleData/AirlineDemoSmall.csv", package = "RevoScaleR")
 data <- RxTextData(source, missingValueString = "M", colInfo = colInfo)

 # When 'distributed' is TRUE distributed data source is created
 distributed <- FALSE
 if (distributed) {
     bigDataDirRoot <- "/share"
     inputDir <- file.path(bigDataDirRoot, "AirlineDemoSmall")
     rxHadoopMakeDir(inputDir)
     rxHadoopCopyFromLocal(source, inputDir)
     hdfsFS <- RxHdfsFileSystem()
     data <- RxTextData(file = inputDir, missingValueString = "M", colInfo = colInfo, fileSystem = hdfsFS)
 }

 # When 'distributed' is TRUE training is distributed
 if (distributed) {
     cc <- rxSetComputeContext(RxSpark())
 } else {
     cc <- rxGetComputeContext()
 }

 ensemble <- rxEnsemble(
     formula = ArrDelay ~ DayOfWeek,
     data = data,
     type = "regression",
     trainers = list(fastTrees(), fastTrees(numTrees = 60), fastTrees(learningRate = 0.1)), # The ensemble will contain three 'rxFastTrees' models
     replace = TRUE # Indicates using a bootstrap sample for each trainer
     )

 # Change the compute context back to previous for scoring
 rxSetComputeContext(cc)

 # Put score and model variables in data frame
 scores <- rxPredict(ensemble, data = data, writeModelVars = TRUE)

 # Plot actual versus predicted values with smoothed line
 rxLinePlot(Score ~ ArrDelay, type = c("p", "smooth"), data = scores)