Ancillary genetic algorithm functions
Built-in functions related to genetic algorithms
These functions are used with the functions argument of the
gafsControl function. More information on the details of these
functions are at http://topepo.github.io/caret/feature-selection-using-genetic-algorithms.html.
Most of the gafs_* functions are based on those from the GA package
by Luca Scrucca. These functions here are small re-writes to work outside of
the GA package.
The objects caretGA, rfGA and treebagGA are example
lists that can be used with the functions argument of
gafsControl.
In the case of caretGA, the ... structure of
gafs passes through to the model fitting routine. As a
consequence, the train function can easily be accessed by
passing important arguments belonging to train to
gafs. See the examples below. By default, using caretGA
will used the resampled performance estimates produced by
train as the internal estimate of fitness.
For rfGA and treebagGA, the randomForest and
bagging functions are used directly (i.e. train is not
used). Arguments to either of these functions can also be passed to them
though the gafs call (see examples below). For these two
functions, the internal fitness is estimated using the out-of-bag estimates
naturally produced by those functions. While faster, this limits the user to
accuracy or Kappa (for classification) and RMSE and R-squared (for
regression).
gafs_initial(vars, popSize, ...) gafs_lrSelection(population, fitness, r = NULL, q = NULL, ...) gafs_spCrossover(population, fitness, parents, ...) gafs_raMutation(population, parent, ...) gafs_nlrSelection(population, fitness, q = 0.25, ...) gafs_rwSelection(population, fitness, ...) gafs_tourSelection(population, fitness, k = 3, ...) gafs_uCrossover(population, parents, ...)
vars |
number of possible predictors |
popSize |
the population size passed into |
... |
not currently used |
population |
a binary matrix of the current subsets with predictors in columns and individuals in rows |
fitness |
a vector of fitness values |
r, q, k |
tuning parameters for the specific selection operator |
parent, parents |
integer(s) for which chromosomes are altered |
The return value depends on the function.
Luca Scrucca, gafs_initial, caretGA, rfGA and
treebagGA by Max Kuhn
Scrucca L (2013). GA: A Package for Genetic Algorithms in R. Journal of Statistical Software, 53(4), 1-37.
pop <- gafs_initial(vars = 10, popSize = 10)
pop
gafs_lrSelection(population = pop, fitness = 1:10)
gafs_spCrossover(population = pop, fitness = 1:10, parents = 1:2)
## Not run:
## Hypothetical examples
lda_ga <- gafs(x = predictors,
y = classes,
gafsControl = gafsControl(functions = caretGA),
## now pass arguments to `train`
method = "lda",
metric = "Accuracy"
trControl = trainControl(method = "cv", classProbs = TRUE))
rf_ga <- gafs(x = predictors,
y = classes,
gafsControl = gafsControl(functions = rfGA),
## these are arguments to `randomForest`
ntree = 1000,
importance = TRUE)
## End(Not run)Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.