sbf: Selection By Filtering (SBF)

Description

Model fitting after applying univariate filters

Usage

sbf(x, ...)
## S3 method for class 'default':
sbf(x, y, sbfControl = sbfControl(), ...)
## S3 method for class 'formula':
sbf(form, data, ..., subset, na.action, contrasts = NULL) 
## S3 method for class 'sbf':
predict(object, newdata = NULL, ...)

Arguments

a data frame containing training data where samples are in rows and features are in columns.

a numeric or factor vector containing the outcome for each sample.

form

A formula of the form y ~ x1 + x2 + ...

data

Data frame from which variables specified in formula are preferentially to be taken.

subset

An index vector specifying the cases to be used in the training sample. (NOTE: If given, this argument must be named.)

na.action

A function to specify the action to be taken if NAs are found. The default action is for the procedure to fail. An alternative is na.omit, which leads to rejection of cases with missing values on any required variable. (NOTE: If given, this argument must

contrasts

a list of contrasts to be used for some or all of the factors appearing as variables in the model formula.

sbfControl

a list of values that define how this function acts. See sbfControl. (NOTE: If given, this argument must be named.)

object

an object of class sbf

newdata

a matrix or data frame of predictors. The object must have non-null column names

...

for sbf: arguments passed to the classification or regression routine (such as randomForest). For predict.sbf: augments cannot be passed to the prediction function

Value

for sbf, an object of class sbf with elements:
predif sbfControl$saveDetails is TRUE, this is a list of predictions for the hold-out samples at each resampling iteration. Otherwise it is NULL
variablesa list of variable names that survived the filter at each resampling iteration
resultsa data frame of results aggregated over the resamples
fitthe final model fit with only the filtered variables
optVariablesthe names of the variables that survived the filter using the training set
callthe function call
controlthe control object
resampleif sbfControl$returnResamp is "all", a data frame of the resampled performance measures. Otherwise, NULL
metricsa character vector of names of the performance measures
dotsa list of optional arguments that were passed in
For predict.sbf, a vector of predictions.

Details

More details on this function can be found at http://caret.r-forge.r-project.org/featureselection.html.

This function can be used to get resampling estimates for models when simple, filter-based feature selection is applied to the training data.

For each iteration of resampling, the predictor variables are univariately filtered prior to modeling. Performance of this approach is estimated using resampling. The same filter and model are then applied to the entire training set and the final model (and final features) are saved.

sbf can be used with "explicit parallelism", where different resamples (e.g. cross-validation group) can be split up and run on multiple machines or processors. By default, sbf will use a single processor on the host machine. As of version 4.99 of this package, the framework used for parallel processing uses the foreach package. To run the resamples in parallel, the code for sbf does not change; prior to the call to sbf, a parallel backend is registered with foreach (see the examples below).

The modeling and filtering techniques are specified in sbfControl. Example functions are given in lmSBF.

Examples

Run this code

data(BloodBrain)

## Use a GAM is the filter, then fit a random forest model
RFwithGAM <- sbf(bbbDescr, logBBB,
                 sbfControl = sbfControl(functions = rfSBF,
                                         verbose = FALSE, 
                                         method = "cv"))
RFwithGAM

predict(RFwithGAM, bbbDescr[1:10,])

## classification example with parallel processing

## library(doMC)

## Note: if the underlying model also uses foreach, the
## number of cores specified above will double (along with
## the memory requirements)
## registerDoMC(cores = 2)

data(mdrr)
mdrrDescr <- mdrrDescr[,-nearZeroVar(mdrrDescr)]
mdrrDescr <- mdrrDescr[, -findCorrelation(cor(mdrrDescr), .8)]

set.seed(1)
filteredNB <- sbf(mdrrDescr, mdrrClass,
                 sbfControl = sbfControl(functions = nbSBF,
                                         verbose = FALSE, 
                                         method = "repeatedcv",
                                         repeats = 5))
confusionMatrix(filteredNB)