基于改进蚱蜢优化算法的特征选择机制

Feature selection mechanism based on improved grasshopper optimization algorithm

针对传统蚱蜢优化算法收敛速度慢、寻优精度低的不足,提出融合学习自动机和折射对立学习的混沌蚱蜢优化算法LRGOA。算法利用完全随机混沌Tent映射进行种群初始化,提高种群遍历性和多样性;利用学习自动机对决定搜索方向的调整系数更新,以均衡全局搜索与局部开发;引入折射对立学习位置更新机制,避免算法陷入局部最优。将LRGOA应用于数据集特征选择问题,设计基于LRGOA的特征选择算法LRGOAFS。选取UCI库数据集对算法的有效性进行验证,证实改进算法可以同步降低特征选择维度和提升数据分类准确率。

For traditional grasshopper optimization algorithm,there exits some insufficiency such as slow convergence speed and low optimization precision.We propose a chaotic grasshopper optimization algorithm LRGOA integrating with learning automata and refraction opposite learning.Our algorithm uses an improved chaos Tent mapping mechanism fusing with complete randomness to realize the population initialization,which can promote the ergodicity and diversity of the population.And,our algorithm use learning automata to determine the update of adjustment coefficient determing the search direction,which can better balanced global search and local development.Our algorithm introduces refraction opposite learning to update the position,avoid a local optimum.LRGOA is applied to solve the feature selection problem in data sets and a feature selection algorithm LRGOAFS based on LRGOA is designed.UCI basedata sets is introduced to test the effective of the algorithm,we prove that the improved algorithm can effectively improve the data classification accuracy with fewer feature dimensions.