基于多目标优化的超盒粒计算分类算法  

The Hyperbox Granular Computing Classification Algorithm Based on Multi-objective Optimization

作　　者：柳春华[1] 刘宏兵[1] 

机构地区：[1]信阳师范学院计算机与信息技术学院,河南信阳464000

出　　处：《信阳师范学院学报（自然科学版）》2014年第1期127-130,共4页Journal of Xinyang Normal University(Natural Science Edition)

基　　金：河南省基础研究与前沿技术项目(132300410421;132300410422);河南省教育厅科学技术研究重点项目(13B520267);河南省教育厅信息技术研究项目(ITE12155);信阳师范学院青年基金项目;信阳师范学院青年骨干教师资助计划

摘　　要：粒的数量和分类错误率是粒计算互相冲突的两个目标,同时最小化这两个目标是不可能的.针对此,构造了多目标优化问题,分别建立分类超盒粒数量和训练错误率两个目标,通过多目标演化算法对该多目标优化问题进行求解,从而产生一系列分类超盒粒集.随机产生初始种群,多目标演化算法通过利用演化操作和反复迭代的方法,得到供用户选取不同性能的解集.Granule number and classification error rate are two conflicting objectives in granular computing , it is impossible to minimize the two objectives simultaneously .The multi-objective optimization including the number of granule number and classification error was formed and solved by multi-objective evolutionary algorithm , and a series of multi-hyperbox granule sets were achieved .The multi-objective evolutionary algorithm obtained the different solution set by initialization of population , evolution operation and iteration method .Users can select the solution according to their requirements .

关 键 词：粒计算 多目标优化 超盒粒 Pareto前端 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]