Genetic Algorithm with Directional Mutation Based on Greedy Strategy for Large-scale 0-1 Knapsack Problems

In view of the lack of efficiency or accuracy of solving large-scale 0-1 knapsack problems by the classic genetic algorithm, a directional mutation operator is designed to reduce the probability of resampling in the search process. Meanwhile an initializing operator and an individual correction operator are added to the algorithm to modify individual after every amendment, both of which are combined with the greedy theory. The proposed algorithm uses truncation selection and longestdistance fitness selection in the crossover, takes elite selection strategy combining with steady-state propagation as a correction operation, and uses the common 0-1 exchange mutation of the binary code, but the mutation probability of each bit of the binary string is adaptively modified. Comparison of experiment results of this algorithm and Active Evolution Genetic Algorithm is given based on the scales of 1000, 2000 and 5000. Experiments proved that the improved algorithm with directional mutation based on greedy theory for solving large-scale 0-1 knapsack problems has high accuracy and high efficiency.