Abstract: In view of the shortcomings of the sparrow search algorithm（SSA） in solving complex optimization problems, such as slow convergence speed, severe population convergence and being easy to fall into local optimum, a multi-strategy improved sparrow search algorithm（MISSA） is proposed. First, the chaotic mapping and reverse learning mechanisms are applied to improve the quality of the initial population. Then, the learning strategy of particle swarm algorithm is introduced to improve the information communication ability of the population and balance the performances of global exploration and local development of the algorithm. Finally, the mutation and cross operations of the differential evolution algorithm are used to enhance the escape power from the local optimal value. Experiments with eight benchmark functions show that the proposed algorithm has better optimization performance and convergence efficiency. Furthermore, the proposed algorithm is applied to optimize the parameters of support vector regression（SVR） model and its effectiveness is demonstrated with five selected UCI datasets.

Key words: sparrow search algorithm, differential mutation, chaotic mapping, opposition-based learning

摘要： 针对麻雀搜索算法在求解复杂优化问题时存在收敛速度慢、种群趋同性严重、易于陷入局部最优等不足，提出一种多策略改进的麻雀搜索算法（multi-strategy improved sparrow search algorithm，MISSA）。通过混沌映射和反向学习机制提高算法初始种群的质量；借鉴粒子群算法的学习策略来提升种群的信息交流能力和兼顾全局勘探与局部开发之间的平衡；融合差分进化算法的变异交叉操作提升算法跳出局部最优值的能力。通过对8个基准测试函数的寻优实验，结果表明改进算法具有更好的优化性能和收敛效率；进一步地，将改进算法应用于优化支持向量回归（support vector regression，SVR）模型的参数，并通过在选定的5个UCI数据集上的实验验证了改进算法的有效性。

关键词: 麻雀搜索算法, 差分变异, 混沌映射, 反向学习