Gene Suppressor: An added phase toward solving large scale optimization problems in genetic algorithm

• Gene Suppressor, proposed as a new add-on phase in GA for attaining self-adaption and repairing.
• This regulates genes dosage and its functional expression with respect to its environment.
• Identifies suppressor genes to perform suppression activity for attaining specific phenotype.
• Allows adjustment by gene adaption and repairing to obtain best solution and improving it.
• Experiment focused on proving single problem but, the buildup model can be easily adopted to other problem that uses MKP as a base.

Genetic algorithm (GA) is a branch of evolutionary algorithm, has proved its effectiveness in solving constrain based complex real world problems in variety of dimensions. The individual phases of GA are the mimic of the basic biological processes and hence the self-adaptability of GA varied in accordance to the adjustable natural processes. In some instances, self-adaptability in GA fails in identifying adaptable genes to form a solution set after recombination, which leads converge toward infeasible solution, sometimes, this, infeasible solution could not be converted into feasible form by means of any of the repairing techniques. In this perspective, Gene Suppressor (GS), a bio-inspired process is being proposed as a new phase after recombination in the classical GA life cycle. This phase works on new individuals generated after recombination to attain self-adaptability by adapting best genes in the environment to regulate chromosomes expression for achieving desired phenotype expression. Repairing in this phase converts infeasible solution into feasible solution by suppressing conflicting gene from the environment. Further, the solution vector expression is improved by inducing best genes expression in the environment within the set of intended constrains. Multiobjective Multiple Knapsack Problems (MMKP), one of the popular NP hard combinatorial problems is being considered as the test-bed for proving the competence of the proposed new phase of GA. The standard MMKP benchmark instances obtained from OR-library [22] are used for the experiments reported in this paper. The outcomes of the proposed method is compared with the existing repairing techniques, where the analyses proved the proficiency of the proposed GS model in terms of better error and convergence rates for all instances.