release_process/andrewma/test-GA1.R

In AndrewM1130/GA: On Multivariate Feature Selection with Genetic Algorithms

library(testthat)
library(assertthat)
source('create_initial_generation.R')
source('more_controllable_data_gen.R')
source('score_fitness_V2.R')
source('elitism.R')
source('test_custom_function.R')

set.seed(1)
data <- generate_data()
generation_matrix <- create_initial_generation(50,50)
gene <- generation_matrix
score_vec <- score_fitness2(gene,data,'AIC')

test <- function(gene,data){
  return(sum(gene))
}

fail <- function(gene,data){
  return(list(gene[1,]))
}

#testing custom user functions
test_that("custom user functions",{
  expect_true(test_user_function(test,gene,data)$message)
  expect_equal(test_user_function(test,gene,data)$error, NA)
  expect_that(test_user_function(fail,gene,data), throws_error())
})



#test the different matrices
test_that("scoring the genes",{
  expect_is(score_fitness2(gene, data, 'AIC'), 'list')
  expect_is(score_fitness2(gene, data, 'R2'), 'list')
  expect_is(score_fitness2(gene, data, custom_function = test), 'list')
})

elite_prop = .001
elite_num <- elite_prop * nrow(data)
elite <- apply_elitism(data, gene, score_vec[[1]], 
                      elite_prop = .01, mutate = 0, metric = 'AIC',
                      family = 'gaussian', custom_function = NULL,
                      fittest = 'high')
  
best <- gene[which.min(score_vec[[2]]), ]

#tests for functionality of elitism

test_that("test elitism",{
  expect_equal(as.numeric(elite), best)
  expect_equal(length(as.numeric(elite)),length(best))
  expect_equal(nrow(elite),elite_num)
})