基于非参数统计的生物启发式优化算法性能评估

摘要

由于生物启发式优化算法性能评估方法存在单一性、片面性且无法进行统一地检验分析问题,从而严重影响了对优化算法性能的深入研究而导致各类优化算法无法准确地应用于解决实际问题。针对这一问题,利用非参数统计检验中典型的 Wilcoxon 符号秩检验和 Quade 检验方法,对5种生物启发式优化算法在36种测试函数条件下的仿真结果进行检验分析。测试结果表明：上述检验方法能够有效地对不同类型的优化算法性能进行分析比较,J ADE 算法相比于其他4种算法,在收敛速度及搜索精度方面表现最优,而 GWO 算法在精度稳定性方面相比于其他4种算法表现出较优的性能,对各类生物启发式优化算法优化性能的评估与比较提供了新的思路。%Aimed at the problems that the biology-inspired optimization algorithms are of oneness,one-sid-edness and fail to check and analyze uniformly the performance evaluation,thus having a strong impact on the intensive study for performance of the optimization algorithms and failing to solve practical problems accurately,two classical nonparametric statistics methods named Wilcoxon Sign Rank test and Quade test are utilized for testing and analyzing the simulation results of five different BOAs under the conditions of thirty-six different test functions.The experimental results show that the two test methods can be used ef-fectively to compare and analyze the optimization performances of different optimization algorithms.JADE algorithm is most superior in convergence speed and search accuracy compared with the other four algo-rithms,whereas,GWO has comparatively superior performance in the aspect of stability compared with other four algorithms.And this provides a new idea for evaluating and comparing the performances of dif-ferent BOAs.