Table 7 Hyperparameters of the genetic algorithm used to perform feature selection
Parameter | Value | Description |
|---|---|---|
Generations | 1,000 | Number of algorithm iterations |
Population size | 300 | Number of candidate solutions (aka individuals) in the population |
Individual representation | Binary | The candidate solutions were represented as a binary array where a value of 1 indicated the presence of a feature and a value of 0 its absence |
Selection | Tournament selection (k = 2) | Stochastically, k individuals are selected from the whole population. From this selection, the individual with the best fitness value is selected for the next generation. This process is repeated until fill established population size |
Elitism | 30 | Individuals who will pass to the next generation without undergoing the selection process |
Mutation | Bit-flip (probability = 0.05) | In each generation, each position of the candidate solution inverts its value according to the specified probability |
Cross-over | One-point (probability = 0.5) | In each generation, two individuals are selected with a certain probability and their information is combined by splitting the solution by a certain cut-off point. The resulting fragments are used to generate the offspring |