Variety and diversity of population are essential for convergence to global optimal in genetic algorithm. In this study, the concepts of fitness distribution, expected and cumulative fitness distribution, reproduction rate and loss of diversity are defined for a sexual selection mechanism, and their performance of this type of selection mechanism is studied theoretically. Then a genetic algorithm based on this selection mechanism and penalty function are utilized for solving multidimensional 0/1 knapsack problems. Computational experiments are conducted on the proposed technique compared with some commonly used selection mechanisms for solving multidimensional 0/1 knapsack problems from the literature.

Key words: Genetic algorithm, multidimensional knapsack problems, penalty function, sexual selection.