A New Evolutionary-Incremental Framework for Feature Selection

Algorithm 1

Standard genetic algorithm.

Input: : recombination probability, : mutation probability

(i) Initialization: is a random initial population

(ii) Evaluation: calculation of fitness value of each solution in the population

(iii) Parent Selection: randomly selection of parents from the population by a given parent selection method (e.g., roulette wheel)

(iv) Recombination: applying a recombination operator (e.g., 1-point crossover) on each pair of parents with the probability

    of to generate new offspring

(v) Mutation: applying a mutation operator (e.g., uniform mutation) on each offspring generated from the previous step

   with the probability of

(vi) Evaluation: calculation of fitness value of offspring

(vii) Survival Selection: randomly selection of individuals from the current population and generated offspring by a given

  survival selection method (e.g., fitness based or age based)

(viii) Termination Condition: if termination condition(s) is satisfied, the algorithm is finished; otherwise go to step (iii).

   When the algorithm is finished, the best solution with the highest fitness value is the output of the algorithm.