Nothing
R/tdmOptsDefaults.r
In TDMR: Tuned Data Mining in R
Defines functions
tdmOptsDefaultsSet
tdmOptsDefaultsFill
Documented in
tdmOptsDefaultsFill
tdmOptsDefaultsSet
######################################################################################
#tdmOptsDefaultsSet:
#
#' Default values for list \code{opts}.
#'
#' Set up and return a list \code{opts} with default settings. The list \code{opts}
#' contains all DM-related settings which are needed by main_<TASK>.
#' \cr\cr
#' For better readability, most elements of \code{opts} are arranged in groups:
#' \tabular{ll}{
#' \code{dir.*} \tab path-related settings \cr
#' \code{READ.*} \tab data-reading-related settings \cr
#' \code{TST.*} \tab resampling-related settings (training, validation and test set, CV) \cr
#' \code{PRE.*} \tab preprocessing parameters \cr
#' \code{SRF.*} \tab several parameters for \code{\link{tdmModSortedRFimport}} \cr
#' \code{MOD.*} \tab general settings for models and model building \cr
#' \code{RF.*} \tab several parameters for model RF (Random Forest) \cr
#' \code{SVM.*} \tab several parameters for model SVM (Support Vector Machines) \cr
#' \code{ADA.*} \tab several parameters for model ADA (AdaBoost) \cr
#' \code{CLS.*} \tab classification-related settings \cr
#' \code{GD.*} \tab settings for the graphic devices \cr
#' }
#'
#' The path-related settings are relative to \code{opts$path}, if it is def'd, else relative to the current dir. \cr
#' Finally, the function \code{\link{tdmOptsDefaultsFill}(opts)} is called to fill in further details, depending on the current
#' settings of \code{opts}.
#'
#' @param opts (optional) the options already set
#' @param path ["."] where to find everything for the DM task.
#'
#' @return a list \code{opts}, with defaults set for all options relevant for a DM task,
#' containing the following elements
#' \item{path}{["."] where to find everything for the DM task}
#' \item{dir.txt}{[data] where to find .txt/.csv files}
#' \item{dir.data}{[data] where to find other data files, including .Rdata }
# \item{dir.Rdata}{[Rdata] -- deprecated, use opts$dir.data -- }
#' \item{dir.output}{[Output] where to put output files}
#' \item{filename}{["default.txt"] the task data}
#' \item{filetest}{[NULL] the test data, only relevant for READ.TstFn!=NULL}
# \item{fileMode}{[FALSE] if =T, write opts$EVALFILE=*_train_eval.csv, *_train.csv.SRF.*.RData file and *_train.log file}
# \item{logFile}{[FALSE] if =T, and if opts$fileMode=T, write log file to *_train.log }
#' \item{data.title}{["Default Data"] title for plots}
#' \item{READ.TXT}{[T] =T: read data from .csv and save as .Rdata, =F: read from .Rdata}
#' \item{READ.NROW}{[-1] read this amount of rows or -1 for 'read all rows'}
#' \item{READ.TrnFn}{ function to be passed into \code{\link{tdmReadDataset}}. Signature: function(opts)
#' returning a data frame. It reads the train-validation data. }
#' \item{READ.TstFn}{ [NULL] function to be passed into \code{\link{tdmReadDataset}}. Signature: function(opts)
#' returning a data frame. It reads a separate test data file. If NULL, this reading step is skipped. }
#' \item{READ.INI}{[TRUE] read the task data initially, i.e. prior to tuning, using \code{\link{tdmReadDataset}} .
#' If =FALSE, the data are read anew in each pass through main_TASK, i.e. in each tuning step (deprecated). }
#' \item{TST.kind}{["rand"] one of the choices from \{"cv","rand","col"\}, see \code{\link{tdmModCreateCVindex}} for details }
#' \item{TST.COL}{["TST.COL"] name of column with train/test/disregard-flag}
#' \item{TST.NFOLD}{[3] number of CV-folds (only for TST.kind=="cv")}
#' \item{TST.valiFrac}{[0.1] set this fraction of the train-validation data aside for validation (only for TST.kind=="rand")}
#' \item{TST.testFrac}{[0.1] set prior to tuning this fraction of data aside for testing (if tdm$umode=="SP_T" and opts$READ.INI==TRUE)
#' or set this fraction of data aside for testing after tuning (if tdm$umode=="RSUB" or =="CV") }
#' \item{TST.trnFrac}{[NULL] train set fraction, if NULL then \code{\link{tdmModCreateCVindex}} will set it to 1 - opts$TST.valiFrac. }
#' \item{TST.SEED}{[NULL] a seed for the random test set selection (\code{\link{tdmRandomSeed}}) and random validation set selection.
#' (\code{\link{tdmClassifyLoop}}). If NULL, use \code{\link{tdmRandomSeed}}. }
#' \item{PRE.PCA}{["none" (default)|"linear"] PCA preprocessing: [don't | do normal PCA (prcomp) ] }
#' \item{PRE.PCA.REPLACE}{[T] =T: replace with the PCA columns the original numerical columns, =F: add the PCA columns }
#' \item{PRE.PCA.npc}{[0] if >0: add monomials of degree 2 from the first PRE.PCA.npc columns (PCs) (only active, if opts$PRE.PCA!="none")}
#' \item{PRE.SFA}{["none" (default)|"2nd"] SFA preprocessing (see package \code{\link[rSFA]{rSFA-package}}: [don't | do ormal SFA with 2nd degree expansion ] }
#' \item{PRE.SFA.REPLACE}{[F] =T: replace the original numerical columns with the SFA columns; =F: add the SFA columns }
#' \item{PRE.SFA.npc}{[0] if >0: add monomials of degree 2 from the first PRE.SFA.npc columns (only acitve, if opts$PRE.SFA!="none") }
#' \item{PRE.SFA.PPRANGE}{[11] number of inputs after SFA preprocessing, only those inputs enter into SFA expansion }
#' \item{PRE.SFA.ODIM}{[5] number of SFA output dimensions (slowest signals) to return }
#' \item{PRE.SFA.doPB}{[T] =F|T: don't | do parametric bootstrap for SFA in case of marginal training data }
#' \item{PRE.SFA.fctPB}{[sfaPBootstrap] the function to call in case of parametric bootstrap, see \code{\link[rSFA]{sfaPBootstrap}}
#' in package \code{\link[rSFA]{rSFA-package}} for its interface description }
#' \item{PRE.allNonVali}{[F] if =T, then use all non-validation data in the training-validation set for PCA or SFA preprocessing.
#' If =F, use only the training set for PCA or SFA processing (only relevant if opts$PRE.PCA!="none" or opts$PRE.SFA!="none"). }
#' \item{PRE.Xpgroup}{[0.99] bind the fraction 1-PRE.Xpgroup in column OTHER (see \code{\link{tdmPreGroupLevels}}) }
#' \item{PRE.MaxLevel}{[32] bind the N-32+1 least frequent cases in column OTHER (see \code{\link{tdmPreGroupLevels}}) }
#' \item{SRF.kind}{["xperc" (default) |"ndrop" |"nkeep" |"none" ] the method used for feature selection, see \code{\link{tdmModSortedRFimport}} }
#' \item{SRF.ndrop}{ [0] how many variables to drop (only relevant if SRF.kind=="ndrop") }
#' \item{SRF.nkeep}{ [NULL] how many variables to keep, NULL="keep all" (only relevant if SRF.kind=="nkeep") }
#' \item{SRF.XPerc}{ [0.95] if >=0, keep that importance percentage, starting with the most important variables (if SRF.kind=="xperc") }
#' \item{SRF.calc}{ [T] =T: calculate importance & save on SRF.file, =F: load from srfFile
#' (srfFile = Output/<confFile>.SRF.Rdata) }
#' \item{SRF.ntree}{ [50] number of RF trees }
#' \item{SRF.samp}{ sampsize for RF in importance estimation.
#' See RF.samp for further info on sampsize. }
#' \item{SRF.verbose}{ [2] }
#' \item{SRF.maxS}{ [40] how many variables to show in plot }
#' \item{SRF.minlsi}{ [1] a lower bound for the length of SRF$input.variables }
#' \item{SRF.method}{ ["RFimp"] }
#' \item{SRF.scale}{ [TRUE] option 'scale' for call importance() in \code{\link{tdmModSortedRFimport}} }
#' \item{MOD.SEED}{[NULL] a seed for the random model initialization (if model is non-deterministic). If NULL, use \code{\link{tdmRandomSeed}}. }
#' \item{MOD.method}{["RF" (default) |"MC.RF" |"SVM" |"NB" ]: use [RF | MetaCost-RF | SVM | Naive Bayes ] in \code{\link{tdmClassify}} \cr
#' ["RF" (default) |"SVM" |"LM" ]: use [RF | SVM | linear model ] in \code{\link{tdmRegress}} }
#' \item{RF.ntree}{[500] }
#' \item{RF.samp}{[1000] sampsize for RF in model training. If RF.samp is a scalar, then it specifies the
#' total size of the sample. For classification, it can also be a vector of length n.class
#' (= # of levels in response variable), then it specifies the size of each strata. The sum
#' of the vector is the total sample size. If NULL, RF.samp will be replaced by 3000 later
#' in tdmModAdjustSampsize*. }
#' \item{RF.mtry}{[NULL] }
#' \item{RF.nodesize}{[1] }
#' \item{RF.OOB}{[TRUE] if =T, return OOB-training set error as tuning measure; if =F, return validation set error }
#' \item{RF.p.all}{[FALSE] }
#' \item{SVM.kernel}{[3] =1: linear, =2: polynomial, =3: RBF, =4: sigmoid}
#' \item{SVM.epsilon}{[0.005] needed only for regression}
#' \item{SVM.gamma}{[0.005] }
#' \item{SVM.coef0}{[0.0] (needed only for opts$SVM.kernel=="polynomial" or =="sigmoid")}
#' \item{SVM.degree}{[3] (needed only for opts$SVM.kernel=="polynomial")}
#' \item{SVM.tolerance}{[0.008] }
#' \item{ADA.coeflearn}{[1] =1: "Breiman", =2: "Freund", =3: "Zhu" as value for boosting(...,coeflearn,...) (AdaBoost) }
#' \item{ADA.mfinal}{[10] number of trees in AdaBoost = mfinal boosting(...,mfinal,...) }
#' \item{ADA.rpart.minsplit}{[20] minimum number of observations in a node in order for a split to be attempted }
#' \item{CLS.cutoff}{[NULL] vote fractions for the classes (vector of length n.class = # of levels in response variable).
#' The class i with maximum ratio (\% votes)/CLS.cutoff[i] wins.
#' If NULL, then each class gets the cutoff 1/n.class (i.e. majority vote wins).
#' The smaller CLS.cutoff[i], the more likely class i will win. }
#' \item{CLS.CLASSWT}{ [NULL] class weights for the n.class classes, e.g. \cr
#' c(A=10,B=20) for a 2-class problem with classes A and B \cr
#' (the higher, the more costly is a misclassification of that real class). It should be a named vector with the same
#' length and names as the levels of the response variable. If no names are given, the levels of the response variables
#' in lexicographical order will be attached in \code{\link{tdmClassify}}. CLS.CLASSWT=NULL for no weights. }
#' \item{CLS.gainmat}{[NULL] (n.class x n.class) gain matrix. If NULL, CLS.gainmat will be set to unit matrix in \code{\link{tdmClassify}} }
#' \item{rgain.type}{["rgain" (default) |"meanCA" |"minCA" ] in case of \code{\link{tdmClassify}}: For
#' classification, the measure \code{Rgain} returned from \code{\link{tdmClassifyLoop}} in
#' \code{result$R_*} is [relative gain (i.e. gain/gainmax) | mean class accuracy | minimum
#' class accuracy | minus Y ]. The goal is to maximize \code{Rgain}. \cr
#' For binary classification there are the additional measures [ "arROC" | "arLIFT"
#' | "arPRE" | "bYouden" ], see 'Value' in \code{\link{tdmModConfmat}}. \cr
#' For regression, the goal is to minimize \code{result$R_*} returned from \code{\link{tdmRegress}}.
#' In this case, possible values are \code{rgain.type} = ["rmae" (default) |"rmse" | "made" ]
#' which stands for [ relative mean absolute error | root mean squared error |
#' mean absolute deviation ]. }
#' \item{ncopies}{[0] if >0, activate \code{\link{tdmParaBootstrap}} in \code{\link{tdmClassify}} }
#' \item{fct.postproc}{[NULL] name of a function with signature \code{(pred, dframe, opts)} where \code{pred} is the prediction of the model on the
#' data frame \code{dframe} and \code{opts} is this list. This function may do some postprocessing on \code{pred} and
#' it returns a (potentially modified) \code{pred}. This function will be called in \code{\link{tdmClassify}} if it is not \code{NULL}. }
#' \item{GD.DEVICE}{["win"] ="win": all graphics to (several) windows (\code{windows} or \code{X11} in package \code{grDevices}) \cr
#' ="rstudio": same as "win", but all graphics go to the RStudio device \cr
#' ="pdf": all graphics to one multi-page PDF \cr
#' ="png": all graphics in separate PNG files in \code{opts$GD.PNGDIR} \cr
#' ="non": no graphics at all \cr
#' This concerns the TDMR graphics, not the SPOT (or other tuner) graphics.
#' If running R from RStudio (if there is a device with name "RStudioGD")
#' then the default "win" is changed to "rstudio" automatically. }
#' \item{GD.RESTART}{[T] =T: restart the graphics device (i.e. close all 'old' windows or re-open
#' multi-page pdf) in each call to \code{\link{tdmClassify}} or \code{\link{tdmRegress}}, resp. \cr
#' =F: leave all windows open (suitable for calls from SPOT) or write more pages in same pdf. }
#' \item{GD.CLOSE}{[T] =T: close graphics device "png", "pdf" at the end of main_*.r (suitable for main_*.r solo) or \cr
#' =F: do not close (suitable for call from tdmStartSpot2, where all windows should remain open) }
#' \item{NRUN}{[2] how many runs with different train & test samples - or - how many CV-runs, if \code{opts$TST.kind}="cv" }
#' \item{APPLY_TIME}{[FALSE] }
#' \item{test2.show}{[FALSE] }
#' \item{test2.string}{["default cutoff"] }
#' \item{VERBOSE}{[2] =2: print much output, =1: less, =0: none}
#'
#' @note The variables opts$PRE.PCA.numericV and opts$PRE.SFA.numericV (string vectors of numeric input columns to be used for PCA or SFA)
#' are not set by \code{\link{tdmOptsDefaultsSet}} or \code{\link{tdmOptsDefaultsFill}}. Either they are supplied by the user or,
#' if NULL, TDMR will set them to \code{input.variables} in \code{\link{tdmClassifyLoop}}, assuming that all columns are numeric.
#----- now obsolete: ----
# If PCA is done, its output \code{pca$numeric.variables} will overwrite \code{opts$PRE.SFA.numericV} (because the numeric variables
# after PCA become the input for SFA).
#'
#' @seealso \code{\link{tdmOptsDefaultsFill}} \code{\link{tdmDefaultsFill}}
#' @author Wolfgang Konen, THK, 2013 - 2018
#' @export
######################################################################################
tdmOptsDefaultsSet <- function(opts=NULL, path=".") {
# --- keep opts a flat list (no other list as members), so that the user can set
# --- individual members before calling setParams(myOpts,defaultOpts())
if (is.null(opts)) {
opts = list()
opts$path <- path
opts$dir.data <- paste(".", "data/", sep="/")
opts$dir.txt <- paste(".", "data/", sep="/")
opts$dir.Rdata <- paste(".", "Rdata/", sep="/")
opts$dir.output <- paste(".", "Output/", sep="/")
opts$filename = "default.txt"
opts$data.title <- "Default Data"
opts$fileMode = FALSE # =T: write opts$EVALFILE=*_train_eval.csv, *_train.csv.SRF.*.RData file and *_train.log file
opts$logFile = FALSE
opts$READ.TXT = TRUE # =T: read data from .csv and save as .Rdata, =F: read from .Rdata
opts$READ.NROW = -1 # [-1] read this amount of rows or -1 for 'read all rows'
opts$READ.INI = TRUE; # read in the task data initially, i.e. prior to tuning
opts$TST.kind <- "rand" # ["cv"|"rand"|"col"] see tdmModCreateCVindex in tdmModelingUtils.r
opts$TST.COL ="TST.COL";# column with train/test/disregard-flag
opts$TST.NFOLD = 3 # number of CV-folds (only for TST.kind=="cv")
opts$TST.valiFrac = 0.10 # set this fraction of data aside for validation (only for TST.kind=="rand")
opts$TST.SEED = NULL # [NULL] a seed for the random test set selection
opts$MOD.SEED = NULL # [NULL] a seed for the random model initialization (if model is non-deterministic)
opts$PRE.PCA = "none" # ["none"|"linear"|"kernel"] PCA preprocessing: [don't | normal pca (prcomp) | kernel pca (kernlab) ]
# (so far we have problems with kernlab in this app --> currently commented out in tdmPreprocUtils.r)
opts$PRE.knum = 0; # [0] if >0 and if PRE.PCA="kernel", take only a subset of PRE.knum records from dset
opts$PRE.PCA.REPLACE=T; # [T] =T: replace the original numerical columns with the PCA columns; =F: add the PCA columns
opts$PRE.PCA.npc <- 0; # [0] if >0: add monomials of degree 2 for the first PRE.PCA.npc columns (PCs)
opts$PRE.SFA = "none" # ["none"|"2nd"] SFA preprocessing: [don't | normal SFA with 2nd degree expansion ]
opts$PRE.SFA.REPLACE=F; # [F] =T: replace the original numerical columns with the SFA columns; =F: add the SFA columns
opts$PRE.SFA.npc <- 0; # [0] if >0: add monomials of degree 2 for the first PRE.SFA.npc columns
opts$PRE.SFA.PPRANGE=11;
opts$PRE.SFA.ODIM=5;
opts$PRE.SFA.doPB=TRUE;
opts$PRE.SFA.fctPB=rSFA::sfaPBootstrap;
opts$PRE.allNonVali=F; # [F] =T: use all non-validation data for PCA or SFA preprocessing (only relevant if opts$PRE.PCA!="none" or opts$PRE.SFA!="none")
opts$PRE.Xpgroup=0.99;
opts$PRE.MaxLevel=32;
opts$SRF.kind = "xperc" # ["xperc"|"ndrop"|"nkeep"|"none"] see tdmModSortedRFimport in tdmModelingUtils.r
opts$SRF.ndrop = 10; # 0..n: how many variables (those with lowest importance) to drop (only for SRF.kind=="ndrop")
opts$SRF.ntree=50; #
opts$SRF.verbose=2; #
opts$SRF.maxS=40; #
opts$SRF.minlsi=1; #
opts$SRF.XPerc = 0.95; # 0.0..1.0: how much of the overall importance to keep
opts$SRF.scale = TRUE; # option 'scale' for call importance() in tdmModSortedRFimport
opts$SRF.calc = TRUE # =T: calculate SRF variable ranking
# =F: reload previously calculated and stored variable ranking
opts$MOD.method="RF"; # ["RF"|"MC.RF"|"SVM"|"NB"]: use [RF| MetaCost-RF| SVM| Naive Bayes] in tdmClassify
# ["RF"|"SVM"|"LM"]: use [RF| SVM| linear model] in tdmRegress
opts$RF.ntree = 500
opts$RF.samp = 1000 # NULL
opts$RF.mtry = NULL
opts$RF.nodesize = 1
opts$RF.OOB = TRUE; # if =T, return OOB-training set error as tuning measure; if =F, return test set error
opts$RF.p.all=FALSE;
opts$SVM.kernel=3; # [3] =1: linear, =2: polynomial, =3: RBF, =4: sigmoid
opts$SVM.cost=1.0;
opts$SVM.epsilon=0.005; # needed only for regression
opts$SVM.gamma=0.005;
opts$SVM.coef0=0.0; # needed only for opts$SVM.kernel=="polynomial" or =="sigmoid"
opts$SVM.tolerance=0.008;
opts$SVM.degree=3; # needed only for opts$SVM.kernel=="polynomial"
opts$ADA.coeflearn=1; # [1] =1: "Breiman", =2: "Freund", =3: "Zhu" as value for boosting(...,coeflearn,...) (AdaBoost)
opts$ADA.mfinal=10; # [10] number of trees in AdaBoost = mfinal boosting(...,mfinal,...)
opts$ADA.rpart.minsplit=20; # minimum number of observations in a node in order for a split to be attempted
opts$CLS.cutoff = NULL # [NULL] vote fractions for the n.class classes. The class i with
# maximum ratio (% votes)/RF.cutoff[i] wins. If NULL, then each
# class gets the cutoff 1/n.class (i.e. majority vote wins).
opts$CLS.CLASSWT = NULL # class weights for the n.class classes,
opts$CLS.gainmat <- NULL# if [NULL], opts$CLS.gainmat will be set to unit matrix in \code{\link{tdmClassify}}
opts$rgain.type="rgain" # ["rgain" (def.)|"meanCA"|"minCA"] for tdmClassify: the measure result$R_* will contain
# relative gain (i.e. gain/gainmax), mean or minimum class accuracy.
# In each case the goal is to maximize the measure.
# ["rmae" (default) |"rmse" ] for regression (tdmRegress). Here the goal is to minimize result$R_*.
opts$ncopies = 0; # if >0, activate tdmParaBootstrap in tdmClassify
opts$DO.POSTPROC = FALSE; # --- deprecated, use opts$fct.postproc=NULL ---
opts$fct.postproc=NULL;
opts$DO.GRAPHICS=T # --- deprecated, use opts$GD.DEVICE="non" ---
opts$GD.DEVICE="win" # ="pdf": all graphics to one multi-page PDF
# ="win": all graphics to (several) windows (X11)
# ="rstudio": all graphics to the *one* RStudio graphics device
# ="non": no graphics at all
# This concerns the TDMR graphics, not the SPOT graphics
opts$GD.RESTART=TRUE # [T] =T: restart the graphics device (i.e. close all 'old' windows
# or re-open multi-page pdf) in each call to tdmClassfiy or tdmRegress, resp.
# =F: leave all windows open (suitable for calls from SPOT)
# or write more pages in same pdf
opts$GD.CLOSE=TRUE # [T] =T: close graphics device "png", "pdf" at the end of main_*.r
# (suitable for main_*.r solo) or =F: do not close (suitable for
# call from tdmStartSpot2, where all windows should remain open)
opts$NRUN = 2 # how many runs with different train & test samples - or -
# how many CV-runs, if opts$TST.kind=="cv"
opts$rep=1; # the number of the repeat (1,...,spotConfig$max.repeats) in case of repeated evocations from tuner
# (needed only internally as private storage for the RNG, see tdmClassify.r
opts$APPLY_TIME=FALSE;
opts$test2.show <- FALSE;
opts$test2.string <- "default cutoff";
opts$VERBOSE=2;
opts$srf = list(); # private storage to pass results back/forth from tdmModSortedRFimport
# opts$srf[["resp"]] holds a data frame with results from/for SRF with response variable "resp".
} #if (is.null(opts))
# DON'T!! - saving the absolute path in elements of opts gives problems when envT is saved on .RData and later loaded
# (perhaps on another computer where the absolute path does not exist) --> this made the example in unbiasedRun.r
# not runnable on CRAN computers, because demoSonar.RData had the path
# dir.data = "C:/Users/wolfgang/Documents/R/win-library/2.15/TDMR/demo02sonar/./data/"
# Instead, use opts$path as separate object when reading / writing files.
#
# if (!is.null(path))
# if (!(path=="./" | length(grep(path,opts$dir.data))==1)) {
# #preceede the following strings with 'path' *only*, if path is not "./" and
# #if opts$dir.data not contains 'path' (first pass through tdmOptsDefaulsSet)
# opts$dir.data <- paste(path,opts$dir.data,sep="")
# opts$dir.txt <- paste(path,opts$dir.txt,sep="")
# opts$dir.Rdata <- paste(path,opts$dir.Rdata,sep="")
# opts$dir.output <- paste(path,opts$dir.output,sep="")
# }
class(opts) <- c("tdmOpts","TDM")
opts <- tdmOptsDefaultsFill(opts);
opts;
}
######################################################################################
# tdmOptsDefaultsFill:
#
#' Fill the current \code{opts}.
#'
#' There is no need to call this function directly, use \code{\link{tdmOptsDefaultsSet}}(opts)
#' instead, which calls tdmOptsDefaultsFill in the end. \cr
#' \code{tdmOptsDefaultsFill} fills the current \code{opts} with further default
#' parameters if they are not yet defined. The defaults may depend on previously
#' defined elements of \code{opts} (e.g. \code{opts$filename}).
#'
#' What is the difference between \code{\link{tdmOptsDefaultsSet}} and \code{\link{tdmOptsDefaultsFill}}? --
#' \code{tdmOptsDefaultsSet} is used for all parameters that do NOT depend on previously defined elements of \code{opts}.
#' \code{tdmOptsDefaultsFill} is used to fill in further \code{opts} elements, if not yet defined, depending on
#' previous settings (e. g. \code{opts$LOGFILE} is derived from \code{opts$filename}).
#'
#' @param opts the options
# -- deprecated -- @param suffix the suffix of \code{opts$filename}. If NULL, take opts$filesuffix (which, if also NULL, is
# -- -- inferred from opts$filename)
#' @return \code{opts}, the extended options, where additional elements, if they are not yet defined, are set as:
#' \item{TST.COL}{["TST.COL"] }
#' \item{PDFFILE}{["*_pic.pdf"] file for multipage graphics in case \code{opts$GD.DEVICE}="pdf" }
#' \item{GD.PNGDIR}{["PNG*"] directory for .png files in case \code{opts$GD.DEVICE}="png" }
#' \item{LOGFILE}{["*.log"] where to log the output }
#' \item{EVALFILE}{["*_eval.csv"] file with evaluation results allEVAL }
#' \item{SRF.samp}{sample size for SRF, derived from \code{RF.samp} }
# \item{SRF.cutoff}{[opts$CLS.cutoff] }
#' \item{rgain.string}{ one out of c("RGain", "MeanCA", "MinCA", "RMAE","RMSE", "MADE", "AreaROC", "AreaLift", "AreaPrRe"),
#' depending on \code{opts$rgain.type} }
#'
#' Here * is the stripped part of \code{opts$filename} (w/o suffix).
#'
#' All files and directories in the above settings are relative to dir \code{opts$dir.output}.
#'
#' @seealso \code{\link{tdmOptsDefaultsSet}}
#' @author Wolfgang Konen, THK
#' @keywords internal
#' @export
######################################################################################
tdmOptsDefaultsFill <- function(opts) { #,suffix=NULL) {
if (is.null(opts)) opts = tdmOptsDefaultsSet();
if (class(opts)[1] != "tdmOpts") stop("Class of object opts is not tdmClass");
filename = opts$filename;
ssp = strsplit(filename,".",fixed=TRUE);
if (length(ssp[[1]])==1) {
filename = paste(filename,".suf",sep=""); # filename has no suffix --> add a dummy suffix ".suf"
ssp = strsplit(filename,".",fixed=TRUE);
}
filesuffix = paste(".",tail(unlist(ssp),1),sep="");
if (is.null(opts$TST.COL)) opts$TST.COL="TST.COL";
# bug fix 05/12: removed the is.null-check from opts$PDFFILE, $GD.PNGDIR, $LOGFILE, $EVALFILE
# (otherwise it might happen that opts$LOGFILE = "default.log"):
opts$PDFFILE=sub(filesuffix,"_pic.pdf",filename)
opts$GD.PNGDIR=paste("PNG",sub(filesuffix,"",filename),"/",sep="");
opts$LOGFILE=sub(filesuffix,".log",filename)
opts$EVALFILE=sub(filesuffix,"_eval.csv",filename) # contains evaluation results allEVAL
if (is.null(opts$rgain.type)) opts$rgain.type="rgain";
#if (is.null(opts$rgain.string)) {
rgainTypeVals = c("rgain","meanCA","minCA","bYouden","rmae","rmse","made","arROC","arLIFT","arPRE");
rgainStringVals=c("RGain","MeanCA","MinCA","balYouden","RMAE","RMSE","MADE","AreaROC","AreaLift","AreaPrRe");
ind = which(opts$rgain.type==rgainTypeVals);
if (length(ind)==0) stop(sprintf("Could not find opts$rgain.type='%s' in c(%s)"
,opts$rgain.type,"'rgain','meanCA','minCA','bYouden','rmae','rmse','made','arROC','arLIFT','arPRE'"));
opts$rgain.string = rgainStringVals[ind];
#}
if (!is.null(opts$DO.GRAPHICS) && is.null(opts$GD.DEVICE))
if (opts$DO.GRAPHICS==F) opts$GD.DEVICE="non"; # DO.GRAPHICS is now (01/2011) deprecated
if (!is.null(opts$GRAPHDEV) && is.null(opts$GD.DEVICE))
opts$GD.DEVICE=opts$GRAPHDEV; # opts$GRAPHDEV is now (12/2011) deprecated
#if("rstudio" %in% names(sessionInfo()$loadedOnly)) { # are we running R from RStudio?
# if (opts$GD.DEVICE=="win") opts$GD.DEVICE="rstudio"; # --- if yes, we have only one graphic device
#}
# unfortunately the above if-clause does no longer (with RStudio0.97, R3.0.2) detect
# whether we are running R from RStudio. Use instead:
if(Sys.getenv("RSTUDIO") =="1") {
if (opts$GD.DEVICE=="win") opts$GD.DEVICE="rstudio";
}
# code which was previously in tdmClassify. Now we put it here and call tdmOptsDefaultsFill from tdmClassify
# (cleaner code, less places where opts-values are set)
#
if (is.null(opts$MOD.method)) opts$MOD.method="RF";
if (is.null(opts$RF.p.all)) opts$RF.p.all=FALSE;
if (is.null(opts$RF.OOB)) opts$RF.OOB=TRUE;
if (is.null(opts$APPLY_TIME)) opts$APPLY_TIME=FALSE;
if (is.null(opts$DO.POSTPROC)) opts$DO.POSTPROC=FALSE; # --- deprecated, use opts$fct.postproc=NULL ---
if (is.null(opts$GD.RESTART)) opts$GD.RESTART=TRUE;
if (is.null(opts$VERBOSE)) opts$VERBOSE=2;
if (is.null(opts$test2.show)) opts$test2.show <- FALSE;
if (is.null(opts$test2.string)) opts$test2.string <- "default cutoff";
# code which was previously in tdmModSortedRFimport. Now we put it here and call tdmOptsDefaultsFill from tdmModSortedRFimport
# (cleaner code, less places where opts-values are set)
#
if (is.null(opts$SRF.samp)) opts$SRF.samp=min(opts$RF.samp,3000); # new 06/2011
if (is.null(opts$SRF.cutoff)) opts$SRF.cutoff=opts$CLS.cutoff; # new 01/2011, might change the default behaviour
if (is.null(opts$SRF.kind)) opts$SRF.kind="xperc";
if (is.null(opts$SRF.XPerc)) opts$SRF.XPerc=0.95;
if (is.null(opts$SRF.calc)) opts$SRF.calc=TRUE;
if (is.null(opts$SRF.ntree)) opts$SRF.ntree=50;
if (is.null(opts$SRF.verbose)) opts$SRF.verbose=2;
if (is.null(opts$SRF.maxS)) opts$SRF.maxS=40;
if (is.null(opts$SRF.minlsi)) opts$SRF.minlsi=1;
if (is.null(opts$SRF.method)) opts$SRF.method="RFimp";
if (is.null(opts$SRF.scale)) opts$SRF.scale = TRUE; # option 'scale' for call importance() in tdmModSortedRFimport
opts;
}
Try the TDMR package in your browser
Any scripts or data that you put into this service are public.
TDMR documentation built on March 3, 2020, 1:06 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions tdmOptsDefaultsSet tdmOptsDefaultsFill
Documented in tdmOptsDefaultsFill tdmOptsDefaultsSet
######################################################################################
#tdmOptsDefaultsSet:
#
#' Default values for list \code{opts}.
#'
#' Set up and return a list \code{opts} with default settings. The list \code{opts}
#' contains all DM-related settings which are needed by main_<TASK>.
#' \cr\cr
#' For better readability, most elements of \code{opts} are arranged in groups:
#' \tabular{ll}{
#' \code{dir.*} \tab path-related settings \cr
#' \code{READ.*} \tab data-reading-related settings \cr
#' \code{TST.*} \tab resampling-related settings (training, validation and test set, CV) \cr
#' \code{PRE.*} \tab preprocessing parameters \cr
#' \code{SRF.*} \tab several parameters for \code{\link{tdmModSortedRFimport}} \cr
#' \code{MOD.*} \tab general settings for models and model building \cr
#' \code{RF.*} \tab several parameters for model RF (Random Forest) \cr
#' \code{SVM.*} \tab several parameters for model SVM (Support Vector Machines) \cr
#' \code{ADA.*} \tab several parameters for model ADA (AdaBoost) \cr
#' \code{CLS.*} \tab classification-related settings \cr
#' \code{GD.*} \tab settings for the graphic devices \cr
#' }
#'
#' The path-related settings are relative to \code{opts$path}, if it is def'd, else relative to the current dir. \cr
#' Finally, the function \code{\link{tdmOptsDefaultsFill}(opts)} is called to fill in further details, depending on the current
#' settings of \code{opts}.
#'
#' @param opts (optional) the options already set
#' @param path ["."] where to find everything for the DM task.
#'
#' @return a list \code{opts}, with defaults set for all options relevant for a DM task,
#' containing the following elements
#' \item{path}{["."] where to find everything for the DM task}
#' \item{dir.txt}{[data] where to find .txt/.csv files}
#' \item{dir.data}{[data] where to find other data files, including .Rdata }
# \item{dir.Rdata}{[Rdata] -- deprecated, use opts$dir.data -- }
#' \item{dir.output}{[Output] where to put output files}
#' \item{filename}{["default.txt"] the task data}
#' \item{filetest}{[NULL] the test data, only relevant for READ.TstFn!=NULL}
# \item{fileMode}{[FALSE] if =T, write opts$EVALFILE=*_train_eval.csv, *_train.csv.SRF.*.RData file and *_train.log file}
# \item{logFile}{[FALSE] if =T, and if opts$fileMode=T, write log file to *_train.log }
#' \item{data.title}{["Default Data"] title for plots}
#' \item{READ.TXT}{[T] =T: read data from .csv and save as .Rdata, =F: read from .Rdata}
#' \item{READ.NROW}{[-1] read this amount of rows or -1 for 'read all rows'}
#' \item{READ.TrnFn}{ function to be passed into \code{\link{tdmReadDataset}}. Signature: function(opts)
#' returning a data frame. It reads the train-validation data. }
#' \item{READ.TstFn}{ [NULL] function to be passed into \code{\link{tdmReadDataset}}. Signature: function(opts)
#' returning a data frame. It reads a separate test data file. If NULL, this reading step is skipped. }
#' \item{READ.INI}{[TRUE] read the task data initially, i.e. prior to tuning, using \code{\link{tdmReadDataset}} .
#' If =FALSE, the data are read anew in each pass through main_TASK, i.e. in each tuning step (deprecated). }
#' \item{TST.kind}{["rand"] one of the choices from \{"cv","rand","col"\}, see \code{\link{tdmModCreateCVindex}} for details }
#' \item{TST.COL}{["TST.COL"] name of column with train/test/disregard-flag}
#' \item{TST.NFOLD}{[3] number of CV-folds (only for TST.kind=="cv")}
#' \item{TST.valiFrac}{[0.1] set this fraction of the train-validation data aside for validation (only for TST.kind=="rand")}
#' \item{TST.testFrac}{[0.1] set prior to tuning this fraction of data aside for testing (if tdm$umode=="SP_T" and opts$READ.INI==TRUE)
#' or set this fraction of data aside for testing after tuning (if tdm$umode=="RSUB" or =="CV") }
#' \item{TST.trnFrac}{[NULL] train set fraction, if NULL then \code{\link{tdmModCreateCVindex}} will set it to 1 - opts$TST.valiFrac. }
#' \item{TST.SEED}{[NULL] a seed for the random test set selection (\code{\link{tdmRandomSeed}}) and random validation set selection.
#' (\code{\link{tdmClassifyLoop}}). If NULL, use \code{\link{tdmRandomSeed}}. }
#' \item{PRE.PCA}{["none" (default)|"linear"] PCA preprocessing: [don't | do normal PCA (prcomp) ] }
#' \item{PRE.PCA.REPLACE}{[T] =T: replace with the PCA columns the original numerical columns, =F: add the PCA columns }
#' \item{PRE.PCA.npc}{[0] if >0: add monomials of degree 2 from the first PRE.PCA.npc columns (PCs) (only active, if opts$PRE.PCA!="none")}
#' \item{PRE.SFA}{["none" (default)|"2nd"] SFA preprocessing (see package \code{\link[rSFA]{rSFA-package}}: [don't | do ormal SFA with 2nd degree expansion ] }
#' \item{PRE.SFA.REPLACE}{[F] =T: replace the original numerical columns with the SFA columns; =F: add the SFA columns }
#' \item{PRE.SFA.npc}{[0] if >0: add monomials of degree 2 from the first PRE.SFA.npc columns (only acitve, if opts$PRE.SFA!="none") }
#' \item{PRE.SFA.PPRANGE}{[11] number of inputs after SFA preprocessing, only those inputs enter into SFA expansion }
#' \item{PRE.SFA.ODIM}{[5] number of SFA output dimensions (slowest signals) to return }
#' \item{PRE.SFA.doPB}{[T] =F|T: don't | do parametric bootstrap for SFA in case of marginal training data }
#' \item{PRE.SFA.fctPB}{[sfaPBootstrap] the function to call in case of parametric bootstrap, see \code{\link[rSFA]{sfaPBootstrap}}
#' in package \code{\link[rSFA]{rSFA-package}} for its interface description }
#' \item{PRE.allNonVali}{[F] if =T, then use all non-validation data in the training-validation set for PCA or SFA preprocessing.
#' If =F, use only the training set for PCA or SFA processing (only relevant if opts$PRE.PCA!="none" or opts$PRE.SFA!="none"). }
#' \item{PRE.Xpgroup}{[0.99] bind the fraction 1-PRE.Xpgroup in column OTHER (see \code{\link{tdmPreGroupLevels}}) }
#' \item{PRE.MaxLevel}{[32] bind the N-32+1 least frequent cases in column OTHER (see \code{\link{tdmPreGroupLevels}}) }
#' \item{SRF.kind}{["xperc" (default) |"ndrop" |"nkeep" |"none" ] the method used for feature selection, see \code{\link{tdmModSortedRFimport}} }
#' \item{SRF.ndrop}{ [0] how many variables to drop (only relevant if SRF.kind=="ndrop") }
#' \item{SRF.nkeep}{ [NULL] how many variables to keep, NULL="keep all" (only relevant if SRF.kind=="nkeep") }
#' \item{SRF.XPerc}{ [0.95] if >=0, keep that importance percentage, starting with the most important variables (if SRF.kind=="xperc") }
#' \item{SRF.calc}{ [T] =T: calculate importance & save on SRF.file, =F: load from srfFile
#' (srfFile = Output/<confFile>.SRF.Rdata) }
#' \item{SRF.ntree}{ [50] number of RF trees }
#' \item{SRF.samp}{ sampsize for RF in importance estimation.
#' See RF.samp for further info on sampsize. }
#' \item{SRF.verbose}{ [2] }
#' \item{SRF.maxS}{ [40] how many variables to show in plot }
#' \item{SRF.minlsi}{ [1] a lower bound for the length of SRF$input.variables }
#' \item{SRF.method}{ ["RFimp"] }
#' \item{SRF.scale}{ [TRUE] option 'scale' for call importance() in \code{\link{tdmModSortedRFimport}} }
#' \item{MOD.SEED}{[NULL] a seed for the random model initialization (if model is non-deterministic). If NULL, use \code{\link{tdmRandomSeed}}. }
#' \item{MOD.method}{["RF" (default) |"MC.RF" |"SVM" |"NB" ]: use [RF | MetaCost-RF | SVM | Naive Bayes ] in \code{\link{tdmClassify}} \cr
#' ["RF" (default) |"SVM" |"LM" ]: use [RF | SVM | linear model ] in \code{\link{tdmRegress}} }
#' \item{RF.ntree}{[500] }
#' \item{RF.samp}{[1000] sampsize for RF in model training. If RF.samp is a scalar, then it specifies the
#' total size of the sample. For classification, it can also be a vector of length n.class
#' (= # of levels in response variable), then it specifies the size of each strata. The sum
#' of the vector is the total sample size. If NULL, RF.samp will be replaced by 3000 later
#' in tdmModAdjustSampsize*. }
#' \item{RF.mtry}{[NULL] }
#' \item{RF.nodesize}{[1] }
#' \item{RF.OOB}{[TRUE] if =T, return OOB-training set error as tuning measure; if =F, return validation set error }
#' \item{RF.p.all}{[FALSE] }
#' \item{SVM.kernel}{[3] =1: linear, =2: polynomial, =3: RBF, =4: sigmoid}
#' \item{SVM.epsilon}{[0.005] needed only for regression}
#' \item{SVM.gamma}{[0.005] }
#' \item{SVM.coef0}{[0.0] (needed only for opts$SVM.kernel=="polynomial" or =="sigmoid")}
#' \item{SVM.degree}{[3] (needed only for opts$SVM.kernel=="polynomial")}
#' \item{SVM.tolerance}{[0.008] }
#' \item{ADA.coeflearn}{[1] =1: "Breiman", =2: "Freund", =3: "Zhu" as value for boosting(...,coeflearn,...) (AdaBoost) }
#' \item{ADA.mfinal}{[10] number of trees in AdaBoost = mfinal boosting(...,mfinal,...) }
#' \item{ADA.rpart.minsplit}{[20] minimum number of observations in a node in order for a split to be attempted }
#' \item{CLS.cutoff}{[NULL] vote fractions for the classes (vector of length n.class = # of levels in response variable).
#' The class i with maximum ratio (\% votes)/CLS.cutoff[i] wins.
#' If NULL, then each class gets the cutoff 1/n.class (i.e. majority vote wins).
#' The smaller CLS.cutoff[i], the more likely class i will win. }
#' \item{CLS.CLASSWT}{ [NULL] class weights for the n.class classes, e.g. \cr
#' c(A=10,B=20) for a 2-class problem with classes A and B \cr
#' (the higher, the more costly is a misclassification of that real class). It should be a named vector with the same
#' length and names as the levels of the response variable. If no names are given, the levels of the response variables
#' in lexicographical order will be attached in \code{\link{tdmClassify}}. CLS.CLASSWT=NULL for no weights. }
#' \item{CLS.gainmat}{[NULL] (n.class x n.class) gain matrix. If NULL, CLS.gainmat will be set to unit matrix in \code{\link{tdmClassify}} }
#' \item{rgain.type}{["rgain" (default) |"meanCA" |"minCA" ] in case of \code{\link{tdmClassify}}: For
#' classification, the measure \code{Rgain} returned from \code{\link{tdmClassifyLoop}} in
#' \code{result$R_*} is [relative gain (i.e. gain/gainmax) | mean class accuracy | minimum
#' class accuracy | minus Y ]. The goal is to maximize \code{Rgain}. \cr
#' For binary classification there are the additional measures [ "arROC" | "arLIFT"
#' | "arPRE" | "bYouden" ], see 'Value' in \code{\link{tdmModConfmat}}. \cr
#' For regression, the goal is to minimize \code{result$R_*} returned from \code{\link{tdmRegress}}.
#' In this case, possible values are \code{rgain.type} = ["rmae" (default) |"rmse" | "made" ]
#' which stands for [ relative mean absolute error | root mean squared error |
#' mean absolute deviation ]. }
#' \item{ncopies}{[0] if >0, activate \code{\link{tdmParaBootstrap}} in \code{\link{tdmClassify}} }
#' \item{fct.postproc}{[NULL] name of a function with signature \code{(pred, dframe, opts)} where \code{pred} is the prediction of the model on the
#' data frame \code{dframe} and \code{opts} is this list. This function may do some postprocessing on \code{pred} and
#' it returns a (potentially modified) \code{pred}. This function will be called in \code{\link{tdmClassify}} if it is not \code{NULL}. }
#' \item{GD.DEVICE}{["win"] ="win": all graphics to (several) windows (\code{windows} or \code{X11} in package \code{grDevices}) \cr
#' ="rstudio": same as "win", but all graphics go to the RStudio device \cr
#' ="pdf": all graphics to one multi-page PDF \cr
#' ="png": all graphics in separate PNG files in \code{opts$GD.PNGDIR} \cr
#' ="non": no graphics at all \cr
#' This concerns the TDMR graphics, not the SPOT (or other tuner) graphics.
#' If running R from RStudio (if there is a device with name "RStudioGD")
#' then the default "win" is changed to "rstudio" automatically. }
#' \item{GD.RESTART}{[T] =T: restart the graphics device (i.e. close all 'old' windows or re-open
#' multi-page pdf) in each call to \code{\link{tdmClassify}} or \code{\link{tdmRegress}}, resp. \cr
#' =F: leave all windows open (suitable for calls from SPOT) or write more pages in same pdf. }
#' \item{GD.CLOSE}{[T] =T: close graphics device "png", "pdf" at the end of main_*.r (suitable for main_*.r solo) or \cr
#' =F: do not close (suitable for call from tdmStartSpot2, where all windows should remain open) }
#' \item{NRUN}{[2] how many runs with different train & test samples - or - how many CV-runs, if \code{opts$TST.kind}="cv" }
#' \item{APPLY_TIME}{[FALSE] }
#' \item{test2.show}{[FALSE] }
#' \item{test2.string}{["default cutoff"] }
#' \item{VERBOSE}{[2] =2: print much output, =1: less, =0: none}
#'
#' @note The variables opts$PRE.PCA.numericV and opts$PRE.SFA.numericV (string vectors of numeric input columns to be used for PCA or SFA)
#' are not set by \code{\link{tdmOptsDefaultsSet}} or \code{\link{tdmOptsDefaultsFill}}. Either they are supplied by the user or,
#' if NULL, TDMR will set them to \code{input.variables} in \code{\link{tdmClassifyLoop}}, assuming that all columns are numeric.
#----- now obsolete: ----
# If PCA is done, its output \code{pca$numeric.variables} will overwrite \code{opts$PRE.SFA.numericV} (because the numeric variables
# after PCA become the input for SFA).
#'
#' @seealso \code{\link{tdmOptsDefaultsFill}} \code{\link{tdmDefaultsFill}}
#' @author Wolfgang Konen, THK, 2013 - 2018
#' @export
######################################################################################
tdmOptsDefaultsSet <- function(opts=NULL, path=".") {
# --- keep opts a flat list (no other list as members), so that the user can set
# --- individual members before calling setParams(myOpts,defaultOpts())
if (is.null(opts)) {
opts = list()
opts$path <- path
opts$dir.data <- paste(".", "data/", sep="/")
opts$dir.txt <- paste(".", "data/", sep="/")
opts$dir.Rdata <- paste(".", "Rdata/", sep="/")
opts$dir.output <- paste(".", "Output/", sep="/")
opts$filename = "default.txt"
opts$data.title <- "Default Data"
opts$fileMode = FALSE # =T: write opts$EVALFILE=*_train_eval.csv, *_train.csv.SRF.*.RData file and *_train.log file
opts$logFile = FALSE
opts$READ.TXT = TRUE # =T: read data from .csv and save as .Rdata, =F: read from .Rdata
opts$READ.NROW = -1 # [-1] read this amount of rows or -1 for 'read all rows'
opts$READ.INI = TRUE; # read in the task data initially, i.e. prior to tuning
opts$TST.kind <- "rand" # ["cv"|"rand"|"col"] see tdmModCreateCVindex in tdmModelingUtils.r
opts$TST.COL ="TST.COL";# column with train/test/disregard-flag
opts$TST.NFOLD = 3 # number of CV-folds (only for TST.kind=="cv")
opts$TST.valiFrac = 0.10 # set this fraction of data aside for validation (only for TST.kind=="rand")
opts$TST.SEED = NULL # [NULL] a seed for the random test set selection
opts$MOD.SEED = NULL # [NULL] a seed for the random model initialization (if model is non-deterministic)
opts$PRE.PCA = "none" # ["none"|"linear"|"kernel"] PCA preprocessing: [don't | normal pca (prcomp) | kernel pca (kernlab) ]
# (so far we have problems with kernlab in this app --> currently commented out in tdmPreprocUtils.r)
opts$PRE.knum = 0; # [0] if >0 and if PRE.PCA="kernel", take only a subset of PRE.knum records from dset
opts$PRE.PCA.REPLACE=T; # [T] =T: replace the original numerical columns with the PCA columns; =F: add the PCA columns
opts$PRE.PCA.npc <- 0; # [0] if >0: add monomials of degree 2 for the first PRE.PCA.npc columns (PCs)
opts$PRE.SFA = "none" # ["none"|"2nd"] SFA preprocessing: [don't | normal SFA with 2nd degree expansion ]
opts$PRE.SFA.REPLACE=F; # [F] =T: replace the original numerical columns with the SFA columns; =F: add the SFA columns
opts$PRE.SFA.npc <- 0; # [0] if >0: add monomials of degree 2 for the first PRE.SFA.npc columns
opts$PRE.SFA.PPRANGE=11;
opts$PRE.SFA.ODIM=5;
opts$PRE.SFA.doPB=TRUE;
opts$PRE.SFA.fctPB=rSFA::sfaPBootstrap;
opts$PRE.allNonVali=F; # [F] =T: use all non-validation data for PCA or SFA preprocessing (only relevant if opts$PRE.PCA!="none" or opts$PRE.SFA!="none")
opts$PRE.Xpgroup=0.99;
opts$PRE.MaxLevel=32;
opts$SRF.kind = "xperc" # ["xperc"|"ndrop"|"nkeep"|"none"] see tdmModSortedRFimport in tdmModelingUtils.r
opts$SRF.ndrop = 10; # 0..n: how many variables (those with lowest importance) to drop (only for SRF.kind=="ndrop")
opts$SRF.ntree=50; #
opts$SRF.verbose=2; #
opts$SRF.maxS=40; #
opts$SRF.minlsi=1; #
opts$SRF.XPerc = 0.95; # 0.0..1.0: how much of the overall importance to keep
opts$SRF.scale = TRUE; # option 'scale' for call importance() in tdmModSortedRFimport
opts$SRF.calc = TRUE # =T: calculate SRF variable ranking
# =F: reload previously calculated and stored variable ranking
opts$MOD.method="RF"; # ["RF"|"MC.RF"|"SVM"|"NB"]: use [RF| MetaCost-RF| SVM| Naive Bayes] in tdmClassify
# ["RF"|"SVM"|"LM"]: use [RF| SVM| linear model] in tdmRegress
opts$RF.ntree = 500
opts$RF.samp = 1000 # NULL
opts$RF.mtry = NULL
opts$RF.nodesize = 1
opts$RF.OOB = TRUE; # if =T, return OOB-training set error as tuning measure; if =F, return test set error
opts$RF.p.all=FALSE;
opts$SVM.kernel=3; # [3] =1: linear, =2: polynomial, =3: RBF, =4: sigmoid
opts$SVM.cost=1.0;
opts$SVM.epsilon=0.005; # needed only for regression
opts$SVM.gamma=0.005;
opts$SVM.coef0=0.0; # needed only for opts$SVM.kernel=="polynomial" or =="sigmoid"
opts$SVM.tolerance=0.008;
opts$SVM.degree=3; # needed only for opts$SVM.kernel=="polynomial"
opts$ADA.coeflearn=1; # [1] =1: "Breiman", =2: "Freund", =3: "Zhu" as value for boosting(...,coeflearn,...) (AdaBoost)
opts$ADA.mfinal=10; # [10] number of trees in AdaBoost = mfinal boosting(...,mfinal,...)
opts$ADA.rpart.minsplit=20; # minimum number of observations in a node in order for a split to be attempted
opts$CLS.cutoff = NULL # [NULL] vote fractions for the n.class classes. The class i with
# maximum ratio (% votes)/RF.cutoff[i] wins. If NULL, then each
# class gets the cutoff 1/n.class (i.e. majority vote wins).
opts$CLS.CLASSWT = NULL # class weights for the n.class classes,
opts$CLS.gainmat <- NULL# if [NULL], opts$CLS.gainmat will be set to unit matrix in \code{\link{tdmClassify}}
opts$rgain.type="rgain" # ["rgain" (def.)|"meanCA"|"minCA"] for tdmClassify: the measure result$R_* will contain
# relative gain (i.e. gain/gainmax), mean or minimum class accuracy.
# In each case the goal is to maximize the measure.
# ["rmae" (default) |"rmse" ] for regression (tdmRegress). Here the goal is to minimize result$R_*.
opts$ncopies = 0; # if >0, activate tdmParaBootstrap in tdmClassify
opts$DO.POSTPROC = FALSE; # --- deprecated, use opts$fct.postproc=NULL ---
opts$fct.postproc=NULL;
opts$DO.GRAPHICS=T # --- deprecated, use opts$GD.DEVICE="non" ---
opts$GD.DEVICE="win" # ="pdf": all graphics to one multi-page PDF
# ="win": all graphics to (several) windows (X11)
# ="rstudio": all graphics to the *one* RStudio graphics device
# ="non": no graphics at all
# This concerns the TDMR graphics, not the SPOT graphics
opts$GD.RESTART=TRUE # [T] =T: restart the graphics device (i.e. close all 'old' windows
# or re-open multi-page pdf) in each call to tdmClassfiy or tdmRegress, resp.
# =F: leave all windows open (suitable for calls from SPOT)
# or write more pages in same pdf
opts$GD.CLOSE=TRUE # [T] =T: close graphics device "png", "pdf" at the end of main_*.r
# (suitable for main_*.r solo) or =F: do not close (suitable for
# call from tdmStartSpot2, where all windows should remain open)
opts$NRUN = 2 # how many runs with different train & test samples - or -
# how many CV-runs, if opts$TST.kind=="cv"
opts$rep=1; # the number of the repeat (1,...,spotConfig$max.repeats) in case of repeated evocations from tuner
# (needed only internally as private storage for the RNG, see tdmClassify.r
opts$APPLY_TIME=FALSE;
opts$test2.show <- FALSE;
opts$test2.string <- "default cutoff";
opts$VERBOSE=2;
opts$srf = list(); # private storage to pass results back/forth from tdmModSortedRFimport
# opts$srf[["resp"]] holds a data frame with results from/for SRF with response variable "resp".
} #if (is.null(opts))
# DON'T!! - saving the absolute path in elements of opts gives problems when envT is saved on .RData and later loaded
# (perhaps on another computer where the absolute path does not exist) --> this made the example in unbiasedRun.r
# not runnable on CRAN computers, because demoSonar.RData had the path
# dir.data = "C:/Users/wolfgang/Documents/R/win-library/2.15/TDMR/demo02sonar/./data/"
# Instead, use opts$path as separate object when reading / writing files.
#
# if (!is.null(path))
# if (!(path=="./" | length(grep(path,opts$dir.data))==1)) {
# #preceede the following strings with 'path' *only*, if path is not "./" and
# #if opts$dir.data not contains 'path' (first pass through tdmOptsDefaulsSet)
# opts$dir.data <- paste(path,opts$dir.data,sep="")
# opts$dir.txt <- paste(path,opts$dir.txt,sep="")
# opts$dir.Rdata <- paste(path,opts$dir.Rdata,sep="")
# opts$dir.output <- paste(path,opts$dir.output,sep="")
# }
class(opts) <- c("tdmOpts","TDM")
opts <- tdmOptsDefaultsFill(opts);
opts;
}
######################################################################################
# tdmOptsDefaultsFill:
#
#' Fill the current \code{opts}.
#'
#' There is no need to call this function directly, use \code{\link{tdmOptsDefaultsSet}}(opts)
#' instead, which calls tdmOptsDefaultsFill in the end. \cr
#' \code{tdmOptsDefaultsFill} fills the current \code{opts} with further default
#' parameters if they are not yet defined. The defaults may depend on previously
#' defined elements of \code{opts} (e.g. \code{opts$filename}).
#'
#' What is the difference between \code{\link{tdmOptsDefaultsSet}} and \code{\link{tdmOptsDefaultsFill}}? --
#' \code{tdmOptsDefaultsSet} is used for all parameters that do NOT depend on previously defined elements of \code{opts}.
#' \code{tdmOptsDefaultsFill} is used to fill in further \code{opts} elements, if not yet defined, depending on
#' previous settings (e. g. \code{opts$LOGFILE} is derived from \code{opts$filename}).
#'
#' @param opts the options
# -- deprecated -- @param suffix the suffix of \code{opts$filename}. If NULL, take opts$filesuffix (which, if also NULL, is
# -- -- inferred from opts$filename)
#' @return \code{opts}, the extended options, where additional elements, if they are not yet defined, are set as:
#' \item{TST.COL}{["TST.COL"] }
#' \item{PDFFILE}{["*_pic.pdf"] file for multipage graphics in case \code{opts$GD.DEVICE}="pdf" }
#' \item{GD.PNGDIR}{["PNG*"] directory for .png files in case \code{opts$GD.DEVICE}="png" }
#' \item{LOGFILE}{["*.log"] where to log the output }
#' \item{EVALFILE}{["*_eval.csv"] file with evaluation results allEVAL }
#' \item{SRF.samp}{sample size for SRF, derived from \code{RF.samp} }
# \item{SRF.cutoff}{[opts$CLS.cutoff] }
#' \item{rgain.string}{ one out of c("RGain", "MeanCA", "MinCA", "RMAE","RMSE", "MADE", "AreaROC", "AreaLift", "AreaPrRe"),
#' depending on \code{opts$rgain.type} }
#'
#' Here * is the stripped part of \code{opts$filename} (w/o suffix).
#'
#' All files and directories in the above settings are relative to dir \code{opts$dir.output}.
#'
#' @seealso \code{\link{tdmOptsDefaultsSet}}
#' @author Wolfgang Konen, THK
#' @keywords internal
#' @export
######################################################################################
tdmOptsDefaultsFill <- function(opts) { #,suffix=NULL) {
if (is.null(opts)) opts = tdmOptsDefaultsSet();
if (class(opts)[1] != "tdmOpts") stop("Class of object opts is not tdmClass");
filename = opts$filename;
ssp = strsplit(filename,".",fixed=TRUE);
if (length(ssp[[1]])==1) {
filename = paste(filename,".suf",sep=""); # filename has no suffix --> add a dummy suffix ".suf"
ssp = strsplit(filename,".",fixed=TRUE);
}
filesuffix = paste(".",tail(unlist(ssp),1),sep="");
if (is.null(opts$TST.COL)) opts$TST.COL="TST.COL";
# bug fix 05/12: removed the is.null-check from opts$PDFFILE, $GD.PNGDIR, $LOGFILE, $EVALFILE
# (otherwise it might happen that opts$LOGFILE = "default.log"):
opts$PDFFILE=sub(filesuffix,"_pic.pdf",filename)
opts$GD.PNGDIR=paste("PNG",sub(filesuffix,"",filename),"/",sep="");
opts$LOGFILE=sub(filesuffix,".log",filename)
opts$EVALFILE=sub(filesuffix,"_eval.csv",filename) # contains evaluation results allEVAL
if (is.null(opts$rgain.type)) opts$rgain.type="rgain";
#if (is.null(opts$rgain.string)) {
rgainTypeVals = c("rgain","meanCA","minCA","bYouden","rmae","rmse","made","arROC","arLIFT","arPRE");
rgainStringVals=c("RGain","MeanCA","MinCA","balYouden","RMAE","RMSE","MADE","AreaROC","AreaLift","AreaPrRe");
ind = which(opts$rgain.type==rgainTypeVals);
if (length(ind)==0) stop(sprintf("Could not find opts$rgain.type='%s' in c(%s)"
,opts$rgain.type,"'rgain','meanCA','minCA','bYouden','rmae','rmse','made','arROC','arLIFT','arPRE'"));
opts$rgain.string = rgainStringVals[ind];
#}
if (!is.null(opts$DO.GRAPHICS) && is.null(opts$GD.DEVICE))
if (opts$DO.GRAPHICS==F) opts$GD.DEVICE="non"; # DO.GRAPHICS is now (01/2011) deprecated
if (!is.null(opts$GRAPHDEV) && is.null(opts$GD.DEVICE))
opts$GD.DEVICE=opts$GRAPHDEV; # opts$GRAPHDEV is now (12/2011) deprecated
#if("rstudio" %in% names(sessionInfo()$loadedOnly)) { # are we running R from RStudio?
# if (opts$GD.DEVICE=="win") opts$GD.DEVICE="rstudio"; # --- if yes, we have only one graphic device
#}
# unfortunately the above if-clause does no longer (with RStudio0.97, R3.0.2) detect
# whether we are running R from RStudio. Use instead:
if(Sys.getenv("RSTUDIO") =="1") {
if (opts$GD.DEVICE=="win") opts$GD.DEVICE="rstudio";
}
# code which was previously in tdmClassify. Now we put it here and call tdmOptsDefaultsFill from tdmClassify
# (cleaner code, less places where opts-values are set)
#
if (is.null(opts$MOD.method)) opts$MOD.method="RF";
if (is.null(opts$RF.p.all)) opts$RF.p.all=FALSE;
if (is.null(opts$RF.OOB)) opts$RF.OOB=TRUE;
if (is.null(opts$APPLY_TIME)) opts$APPLY_TIME=FALSE;
if (is.null(opts$DO.POSTPROC)) opts$DO.POSTPROC=FALSE; # --- deprecated, use opts$fct.postproc=NULL ---
if (is.null(opts$GD.RESTART)) opts$GD.RESTART=TRUE;
if (is.null(opts$VERBOSE)) opts$VERBOSE=2;
if (is.null(opts$test2.show)) opts$test2.show <- FALSE;
if (is.null(opts$test2.string)) opts$test2.string <- "default cutoff";
# code which was previously in tdmModSortedRFimport. Now we put it here and call tdmOptsDefaultsFill from tdmModSortedRFimport
# (cleaner code, less places where opts-values are set)
#
if (is.null(opts$SRF.samp)) opts$SRF.samp=min(opts$RF.samp,3000); # new 06/2011
if (is.null(opts$SRF.cutoff)) opts$SRF.cutoff=opts$CLS.cutoff; # new 01/2011, might change the default behaviour
if (is.null(opts$SRF.kind)) opts$SRF.kind="xperc";
if (is.null(opts$SRF.XPerc)) opts$SRF.XPerc=0.95;
if (is.null(opts$SRF.calc)) opts$SRF.calc=TRUE;
if (is.null(opts$SRF.ntree)) opts$SRF.ntree=50;
if (is.null(opts$SRF.verbose)) opts$SRF.verbose=2;
if (is.null(opts$SRF.maxS)) opts$SRF.maxS=40;
if (is.null(opts$SRF.minlsi)) opts$SRF.minlsi=1;
if (is.null(opts$SRF.method)) opts$SRF.method="RFimp";
if (is.null(opts$SRF.scale)) opts$SRF.scale = TRUE; # option 'scale' for call importance() in tdmModSortedRFimport
opts;
}
Try the TDMR package in your browser
Any scripts or data that you put into this service are public.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.