采用渐变与突变机制的反向人工蜂群算法

采用渐变与突变机制的反向人工蜂群算法 An Opposition-Based Artificial Bee Colony Algorithm Using the Mechanism of Gradual and Sudden Change 胡玉荣 HU Yurong1 1 2 first-author 丁立新 DING Lixin1 1 2 谢大同 XIE Datong1 1 2 汪慎文 WANG Shenwen1 1 2 郭肇禄 GUO Zhaolu1 1 2 谢承旺 XIE Chengwang3 3 武汉大学软件工程国家重点试验室 武汉大学软件工程国家重点试验室 State Key Laboratory of Software Engineering,Wuhan University State Key Laboratory of Software Engineering,Wuhan University 武汉大学计算机学院 武汉大学计算机学院 Computer School,Wuhan University Computer School,Wuhan University 华东交通大学软件学院 华东交通大学软件学院 School of Software,East China Jiaotong University School of Software,East China Jiaotong University 为克服人工蜂群算法容易陷入局部最优且后期收敛速度较慢的缺点,提出一种基于渐变与突变机制的反向人工蜂群算法并用于特征选择.采用反向学习策略,为每个初始解产生对应的反向解,并从所有解中选择最优的解构成初始种群,加快了收敛速度.引入渐变与突变机制,将个体按适应度大小分为渐变个体和突变个体,对它们采用不同的邻域搜索方法,避免了陷入局部最优.对比实验表明,新算法比其他特征选择算法能够得到更好的特征子集且具有更快的收敛速度. An opposition-based artificial bee colony algorithm,which uses the mechanism of gradual and sudden change,is proposed to overcome the limitations of premature convergence and stagnation.Adopting the strategy of opposition-based learning,new solutions are generated from the initial solutions,and the initial population is formed by selecting the best solutions,and thus it can speed up the convergence.Meanwhile,the mechanism of gradual and sudden change is introduced to divide the individuals into gradual ones and sudden ones by the fitness.Additionally,different neighborhood search methods aim at different individuals,and thus avoid falling into local optimum.Compared with other algorithms,the experimental results show that the new algorithm can obtain better feature subsets and converge speed. 人工蜂群 特征选择 粗糙集 反向学习 渐变与突变 artificial bee colony feature selection rough set opposition-based learning gradual and sudden change 10.14188/j.1671-8836.2013.02.003 O224;TP18 国家自然科学基金（60975050,61165004）;江西省自然科学基金（20114BAB201025）;福建省自然科学基金（2012J01248）资助项目 2013-02-01 2021-04-01 2 59