In this work it is investigated the use of an island-inspired genetic algorithm with adaptive parameters applied to the Multiple Knapsack Problem. It is known that Knapsack problems represent an important class of problems in practice. In the proposed approach, each individual of the GA population is considered as an island. Also, a method to adapt the control parameters (crossover and mutation rates) is applied. It is known that the optimum values of the control parameters can change over the optimization process, directly influencing the efficiency of the method. In the experiments, 11 benchmarks were used to assess the performance of the proposed methodology. Encouraging results were obtained.