metaheuristic optimization of preprocessing combinations

Share:

Description

metaheuristic optimization of preprocessing combinations

Usage

1
2
3
4
metaheur(gridclassobject, startgrid = 0, startnum = 1, iterations = 10,
  taboolistlength = 1, initialtemperature = 0.01, tempconst = 0.01,
  reheat = 0.01, nholdout = 2, late = 0, stopcond = 1,
  stopvalue = 0.99, deltafive = 0.05, model = "rpart", cores = 1)

Arguments

gridclassobject

(GridClass) created by setgrid function in preprocomb package

startgrid

(integer) 0 random restart (default), 1 grid restart

startnum

(integer) number of restarts

iterations

(integer) number of iterations done for a restart, defaults to 10

taboolistlength

(integer) number of previous solution that can not be revisited, must be 1 or more

initialtemperature

(numeric) initial propability for acccepting an inferior candidate, between 0 and 1

tempconst

(numeric) multiplier for decreasing temperature on each iteration

reheat

(numeric) propability of increasing temperature on each iteration

nholdout

(integer) number of holdout rounds, defaults to 2

late

(integer) location of previous best solution a candidate is compared to, defaults to 0 for last

stopcond

(integer) type of stopping condition in addition to iterations, default to 1 for threshold, 2 for convergence

stopvalue

(numeric) threshold for stopping, defaults to 0.99

deltafive

(numeric) convergence criteria for last five iterations, defaults to 0.05

model

(character) caret name of predictive model, defaults to "rpart"

cores

(integer) number of cores used in computation of classification accuracy holdout rounds, defaults to 1

Examples

1
2
## result <- metaheur(examplegrid, startnum=2, nholdout=2, cores=2)
## getbestheur(result)