垂直分布多决策表下基于条件信息熵的近似约简  

Approximate reduction based on conditional information entropy over vertically partitioned multi-decision table

作　　者：杨明[1] 杨萍[1] 

机构地区：[1]南京师范大学数学与计算机科学学院,南京210097

出　　处：《控制与决策》2008年第10期1103-1108,共6页Control and Decision

基　　金：国家自然科学基金项目(40771163);江苏省自然科学基金项目(BK2005135);江苏省高校自然科学研究项目(05KJB5200665)

摘　　要：目前粗糙集理论研究主要针对单个决策表,而有关分布式环境下的核求解和属性约简研究的报道不多,为此提出垂直分布多决策表下基于条件信息熵的近似约简算法.该算法在各局部站点并行求相应的条件信息熵,并通过传送部分等价类的策略,可有效降低通讯代价,提高垂直分布多决策表下基于条件信息熵的近似约简效率.算法分析和实验结果表明,所提出的算法是有效可行的.At present, the research based on rough set theory aims mainly at a single decision table, but little work has been done for computation of a core and attribute reduction in distributed environment. Therefore, this paper introduces an approximate reduction algorithm based on conditional information entropy for vertically partitioned multi-decision table. The algorithm computes each corresponding conditional information entropy in parallel at every sites. By employing a data transmitted strategy, the communication cost is efficiently reduced and the performance of the algorithm for approximated reduction based on conditional information entropy over vertically partitioned multi- decision table can be efficiently improved. Algorithm analysis and experimental results show the effectiveness and feasibility of the proposed algorithm.

关 键 词：粗糙集 条件信息熵 全局属性核 局部属性核 近似约简 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]