culebra.fitness_function.feature_selection module

Feature selection related fitness functions.

This sub-module provides several fitness functions to solve feature selecion problems:

  • NumFeats: Dummy single-objective function that minimizes the number of selected features from a Dataset.

  • FeatsProportion: Dummy single-objective function that minimizes the number of selected features from a Dataset. The difference with NumFeats is just that FeatsProportion returns a normalized number in [0, 1].

  • KappaIndex: Single-objective function that maximizes the Kohen’s Kappa index for classification problems.

  • Accuracy: Single-objective function that maximizes the Accuracy for classification problems.

  • KappaNumFeats: Bi-objective function composed by the two former functions. It tries to both maximize the Kohen’s Kappa index and minimize the number of features that a solution has selected.

  • AccuracyNumFeats: Bi-objective function composed by the two former functions. It tries to both maximize the Accuracy and minimize the number of features that a solution has selected.

  • KappaFeatsProp: Bi-objective function composed by the two former functions. It tries to both maximize the Kohen’s Kappa index and minimize the proportion of features that a solution has selected.

  • AccuracyFeatsProp: Bi-objective function composed by the two former functions. It tries to both maximize the Accuracy and minimize the proportion of features that a solution has selected.