Abstract:

The backtracking search optimization algorithm (BSA) is a novel evolution algorithm. However, the BSA has the problem of low convergence speed as the same as the other evolution algorithms. Aiming at this problem, an improved BSA with the comprehensive learning strategy is proposed based on detailed analysis of BSA. The strategy is used for making full use of the better solutions that the population obtains. Firstly, the global best learning equation is proposed and the random evolution equation is introduced in the strategy. They are chosen randomly so as to improve the convergence speed and precision of the improved algorithm. Secondly, in order to control the search direction, the global best search equation is proposed in the strategy so as to reach the global best solution as fast as possible. Finally, 20 complex benchmarks and other three popular algorithms are compared to illustrate the superiority of BSA with comprehensive learning strategy. The experimental results and the Wilcoxon signed ranks test results show that the BSA with comprehensive learning strategy outperformed the other three algorithms in terms of convergence speed and precision.