gep_train: Train a GEP model

In profyliu/gepR: Composite linear-nonlinear regression using GEP

Train a GEP model

Description

Fit a composite linear-nonlinear regression model using Gene Expression Programming. The gepR.dll must be present in the work directory and loaded using the gep_load_dll function.

Usage

gep_train(y, x, px1 = 0.4, px2 = 0.1, pm = 0.3, maxiter = 1000,
  headlen = 5, popsize = 100, eliterate = 0.1, goal = 0.95,
  rseed = 8888, nthreads = 4, verbose = 1, fit_method = 0,
  maxpass = 3, sol_file = "gep_sol.dat")

Arguments

y	numeric vector containing the response variable
x	numeric dataframe containing the independent variables
px1	one-point crossover rate, between 0 and 1
px2	two-point crossover rate, between 0 and 1
pm	mutation rate, between 0 and 1
maxiter	maximum number of iteration (generations) per round
headlen	head length of a gene
popsize	population size (number of individuals per generation), default 100. This is where the OpenMP "parallel for" takes effect. So it is preferred to be an integer multiple of nthreads.
eliterate	elite rate, the proportion in the population to survive to the next generation, between 0 and 1
goal	the targeted R-square value, between 0 and 1, default 0.95. The program stops when R-square achieves this value.
rseed	random seed to be used in the C program, integer value
nthreads	number of parallel threads to be used in computation, ideally set to equal to the number of cores available
verbose	verbose leve, 0 to 2, 0 terse, 2 verbose
fit_method	0: regression, 1: classification, default 0. Only 0 is implemented in this version.
sol_file	character string of the path and name of the file where the GEP model is to be saved.

Value

sol_file

Examples

gep_load_dll()
gepmod <- gep_train(elecdemand[,1], elecdemand[,2:4], nthreads=20)
predval <- gep_score(elecdemand[,2:4], gepmod)
gep_unload_dll()

profyliu/gepR documentation built on Sept. 3, 2023, 8:52 p.m.