In the thesis, a dominance mechanism is proposed and two modified genetic operators are introduced using diploid. Genetic Algorithms(GAs) have been successfully applied to various problems and generally accepted to be effective and robust for a wide range of problems. But when a GA fails to maintain population diversity, it suffers from premature convergence which often ends up converged to a local optima. Diploid GAs have a few advantages over haploid GAs mainly maintaining population diversity and giving the effect of long-term distributed memory on the population. The outline of the proposed diploid GA is as follows: phenotype is determined for each individual, the dominance mechanism uses the fitness value of each chromosome in the individual and maps the one with higher fitness value to its phenotype (Winner-take-all). Dominance mutation operator is applied for each individual which changes the dominance randomly with a certain probability. Crossover is applied for the selected individuals, mating the dominant pair and the recessive pair to produce offsprings for the successive generation. Repeat the procedure until the termination condition. 5 experiments were performed to compare the proposed method with the previous works. The result of the oscillating blind 0/1 knapsack problem shows that it gives more stable long-term distributed memory effect using diploid and the result of dynamic set partitioning problem as well as the other results shows that the proposed method enhances the ability of maintaining population diversity thus preventing premature convergence effectively.