tests/testthat/testSetFSSearchAlgorithmInterface.R

In FSinR: Feature Selection

context("Set feature selection search algorithm interface")
algorithms <- list(whaleOptimization, tabu, sequentialForwardSelection, sequentialBackwardSelection, sequentialFloatingForwardSelection, sequentialFloatingBackwardSelection, simulatedAnnealing, LasVegas, hillClimbing, geneticAlgorithm, antColony, breadthFirst, deepFirst)
for (algorithm in algorithms) {
  result <- algorithm()(iris, 'Species', giniIndex())
  test_that(paste(attr(algorithm(),'name')," output required fields"), {
    expect_true(!is.null(result$bestFeatures))
    expect_true(!is.null(result$bestFitness))
  })
  test_that(paste(attr(algorithm(),'name')," values are correct"), {
    expect_true(all(is.element(result$bestFeatures, c(0,1))))
    expect_true(all(result$bestFeatures <= 1))
    expect_true(all(result$bestFeatures >= 0))
  })
  test_that(paste(attr(algorithm(),'name')," columns number are correct"), {
    expect_equal(ncol(result$bestFeatures), ncol(iris) - 1)
  })
  test_that(paste(attr(algorithm(),'name')," stops if an individual measure is used"), {
    expect_error(algorithm()(iris, 'Species', relief()), 'Only feature set measures can be used')
  })
  test_that(paste(attr(algorithm(), 'name'), " fitness value is in correct range"), {
    eps <- if (capabilities("long.double")) {
      expect_true(all(result$bestFitness <= 1))
      expect_true(all(result$bestFitness >= 0))
    } else {
      expect_true(all(result$bestFitness <= 1.1))
      expect_true(all(result$bestFitness >= -0.1))
    }
  })
}