feat.rank.re: Wrapper for Resampling Based Feature Ranking

Description Usage Arguments Details Value Author(s) See Also Examples

View source: R/ftrank.R

Description

Wrapper for performing feature ranking method on multiple subsets of the original data. At each iteration, only the training samples defined in tr.idx (or optionally pars only) are used to rank the variables. Features rank, order and saliency indicators calculated on the whole data are also given in the output. As for accest this function allows the use of multiple processors as long as the cluster has been set up with the snow package. Data input can be in the form of data matrix + class vector, following the classic formula type or derived from dat.sel1.

Usage

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
  feat.rank.re(...)
  
  ## Default S3 method:
feat.rank.re(x,y,method,pars = valipars(),tr.idx=NULL,clmpi=NULL, seed=NULL, ...)
  
  ## S3 method for class 'formula'
feat.rank.re(formula, data = NULL, ...)

 ## S3 method for class 'dlist'
feat.rank.re(dlist,  method, pars = NULL, tr.idx = NULL, ...)

Arguments

formula

A formula of the form groups ~ x1 + x2 + ... That is, the response is the grouping factor and the right hand side specifies the (non-factor) discriminators.

x

A matrix or data frame containing the explanatory variables.

dlist

A matrix or data frame containing the explanatory variables if no formula is given as the principal argument.

data

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

y

A factor specifying the class for each observation.

method

Feature ranking method to be used. See fs.techniques for details.

pars

A list of resampling scheme or validation method such as Leave-one-out cross-validation, Cross-validation, Bootstrap and Randomised validation (holdout). See valipars for details.

tr.idx

User defined index of training samples of type trainind. Generated by trainind if tr.idx=NULL.

clmpi

snow cluster information

seed

Seed.

...

Additional parameters to be passed to method. See fs.techniques for details.

Details

The structure of the feat.rank.re object is as follows:

method

Feature ranking method used.

fs.rank

A vector of final feature ranking scores.

fs.order

A vector of final feature order from best to worst.

fs.stats

A vector of means of statistics or measurements in all computation.

rank.list

Feature rank list of all computation.

order.list

Feature order list of all computation.

pars

Resampling parameters.

tr.idx

Index of training samples.

pars.min

Condensed form of the resampling strategy for output purposes.

cl.task

Condensed form of the classification task.

all

Feature ranking object originated from the overall dataset.

Value

feat.rank.re object.

Author(s)

David Enot [email protected]

See Also

valipars, ftrank.agg, fs.techniques

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
## load abr1
data(abr1)
y   <- factor(abr1$fact$class)
x <- preproc(abr1$pos , y=y, method=c("log10","TICnorm"),add=1)[,110:500]  
## Select classes 1 and 2
pars   <- valipars(sampling="boot",niter=2,nreps=5)
dat <- dat.sel1(x, y, pwise=c("1","2"),mclass=NULL,pars=pars)

## multiple rankings using AUC
z      <- feat.rank.re(dat[[1]],method="fs.auc")

## print content of z
names(z)
               

wilsontom/FIEmspro documentation built on Feb. 19, 2018, 9:03 a.m.