A Multi-strategy Shuffled Frog-leaping Algorithm for Numerical Optimization

Abstract

Abstract This study proposes a multi-strategy shuffled frog-leaping algorithm for numerical optimization (MSSFLA), which combines the merits of a frog-leaping step rule, the crossover operator, and a novel recursive programming. First, the frog-leaping step rule depends on the level of attractive effect between the worst frog and other frogs in a memeplex, which utilizes the advantages of frogs around the worst frog, making the worst frog more conducive to the evolution direction of the whole population. Second, the crossover operator of the genetic algorithm is used for yielding new frogs based on the best and worst individual frog instead of the random mechanism in the original shuffled frog-leaping algorithm (SFLA). The crossover operation aims to enhance population diversity and conduciveness to the memetic evolution of each memeplex. Finally, recursive programming is presented to store the results of preceding attempts as basis for the computation of those that succeed, which will help save a large number of repeated computing resources in a local search. Experiment results show that MSSFLA has better performance than other algorithms on the convergence and searching effectivity. Therefore, it can be considered as a more competitive improved algorithm for SFLA on the efficiency and accuracy of the best solution.