摘要
为了克服传统粗糙集属性约简方法求解效率不高,且难以搜索出满足用户需求的最优属性约简集的问题,提出了一种属性序约简优化算法。该算法基于决策表的完全属性-值空间树结构,在属性约简空间自适应构造小生境超球面邻域半径,并进行约简树的生成、剪枝、约简及动态优化等,快速找到满足用户需求的最优属性序约简集。相关仿真实验表明该算法在保证收敛速度的同时具有较强的属性约简优化性能,是一种能满足用户需求的高效属性序约简算法。
In order to overcome the poor efficiency of traditional attribute reduction algorithms and the difficulty in searching the optimization attribute reduct set for the user-oriented,a novel attribute order reduct algorithm is proposed here.Based on the attribute-value space tree structure in the decision table,the algorithm can construct the adaptive niche neighborhood radius in the super-sphere,and an attribute reduct tree is constructed,pruned,reduced and optimized.So the global optimization attribute order reduct for the user-oriented is obtained quickly.Experimental results demonstrate the proposed algorithm is better in both convergence efficiency and attribute reduct.So it is efficient to the minimal attribute order reduct for the user-oriented.
出处
《南京理工大学学报》
EI
CAS
CSCD
北大核心
2012年第1期37-42,共6页
Journal of Nanjing University of Science and Technology
基金
国家'863'计划资助项目(2006AA12A106)
国家自然科学基金(61171132)
江苏省普通高校研究生科研创新计划资助项目(CXZZ11_0219)
南通市科技计划项目(BK2011062)
苏州大学江苏省计算机信息处理技术重点实验室开放课题(KJS1023)
南通大学自然科学基金(10Z033)
关键词
小生境邻域半径
属性约简
属性序
完全属性-值空间树
niche neighborhood radius
attribute reduct
attribute order
perfect attribute-value space tree
