等价划分与蚁群优化的属性约简改进策略.pdfVIP

2011 年 12 月 北京邮电大学学报 Dec.2011 第34 卷第6 期 loumal of Beijing University of Posts and Telecommunications Vo l. 34 No.6 文章编号: 1007-5321 (2011 ) 06 -D055-04 等价划分与蚁群优化的属性约简改进策略 王慧1 ， 2 王京1 (1 北京科技大学信息工程学院，北京 100083: 2. 中国人民公安大学信息安全工程系，北京 100038 ) 摘要:为降低经典信息恼属性约简算法的时间复杂度，在论证信息恼属性约简与论域划分细化约简等价的基础上， 提出将蚁群并行优化处理机制引人划分细化约简过程，在蚁群搜索过程中使用体现属性约简特点的状态转移和信 息素更新策略.通过复杂性分析和实例验证，新算法可有效避免蚁群搜索的盲目性，并在较小选代规模下快速获得 约简集，更适于大容量数据表的处理. 关键词:数据挖掘;蚁群优化;等价划分;属性约简 中图分类号: TN929.53 文献标志码:A An Improvement Strategy of Attribute Reduction on Partition and Ant Colony Optimization WANG Hui 1,2 , WANG Jing 1 (1. Schoo1 of Information Engineering , University of Science and Technology , Beijing 100083 , China: 2. Department of Information Security Engineeri吨， Chinese People s Public Security University , Beijing 100038 , China) Abstract: To decrease the time complexity of the attribute reduction algorithm based on information en- tropy , a concurrent processing of ants colony optimization is introduced into partition based on the equiva- lence of attribute reduction between partition and information entropy. In the ant searching process the strategies of state transfer and pheromone update reflect the characteristics of the attribute reduction. Analysis of this algorithm s complexity and simulation show that the blindness of the ant searching can be avoided and the size of the iteration is reduced to make the reduction faste r. For large data base , it is more suitable. Key words: data mining; ant colony opti日lÏzation; partition; attribute reduction 属性约简常作为数据挖掘过程中必要的前序预