regressoR.batchLearn: Apply the Regression Learning to a Directory of Data and...

In thomasWeise/regressoR: An R Package for Fitting Models to 2-Dimensional (x-y) Data

Description

Recursively, a source directory is traversed and all files matching to a selector regular expression are picked up, loaded with a loader, and then modelled by the regressor. The resulting models are stored in a destination folder in a structure mirroring the source folder.

This method uses regressoR.learnForExport to learn the models and stores them into files using saveRDS to store them. They can later be read using regressoR.loadResult.

Usage

regressoR.batchLearn(source = getwd(), destination = file.path(source,
  "../models"), loader = function(file) read.csv(file, sep = "\t", header =
  FALSE)[c(1, 2)], selector = path.extensionRegExp("txt"),
  check.directory = NULL, learn.single = TRUE, learn.all = FALSE,
  learners = regressoR.defaultLearners(), representations = function(x, y)
  Transformation.applyDefault2D(x = x, y = y, addIdentity = TRUE),
  metricGenerator = RegressionQualityMetric.default,
  suffix.single = "_single.model", suffix.all = "_all.model", q = 0.75,
  includeMetric = TRUE, cores = 1L, logging = if (cores <= 1L) {     TRUE
  } else {     file.path(destination, "log.txt") }, returnResults = FALSE,
  skipExisting = (!returnResults))

Arguments

source	the source directory, which is recursively searched for files with data to be modeled
destination	the destination folder, will be created if not existing
loader	a loader function which accepts a vector of paths and is supposed to return an n*2 matrix where the first column contains the x values and the second column the y values to model
selector	a regular expression against which file names are matched. Only matching files are considered.
check.directory	a function receiving a root folder and the path under root and decides whether this directory path should be recursively investigated (be default always TRUE)
learn.single	should every single file matching to the selector be modeled separately ?
learn.all	should all the files in one directory combined and modeled at once?
learners	the model learners to be applied
representations	a function which can transform a x/y dataset into a set of transformed data instances
metricGenerator	the learning quality metric generator
suffix.single	the suffix to append to the files containing the single models
suffix.all	the suffix to be applied to the files containing the models of all files in a folder
q	the modelling quality parameter
includeMetric	should the metric used for learning be stored in the files
cores	the number of cores to use (>1L leads to parallel execution)
logging	should progress information be printed: either TRUE for printing to the console via print, FALSE for no logging, or a path to a file receiving logging information
returnResults	should we return the computed results or not?
skipExisting	should already existing models (resulting from a previous, incomplete execution) simply be skipped or overwritting

See Also

regressoR.learnForExport

regressoR.loadResult

thomasWeise/regressoR documentation built on May 9, 2019, 8:12 p.m.