南京大学学报(自然科学版) ›› 2011, Vol. 47 ›› Issue (4): 383390.
谢娟英**,李楠1,2,乔子茵1
Xie Juan一Ying1,Li Nan1,2,Qiao Zi-Rui1
摘要: 针对不完整决策系统属性约简算法时间复杂度较高问题,基于正域不变条件卜,决策系统分类能力保持不变原则,提出不完整决策系统前向顺序特征选择算法.该算法从约简集为空集开始,根据
在约简集合中加入各属性后对正域影响程度大小将属性降序排列,采用顺序前向搜索,选择当前最佳特征加入特征约简集合,确定最佳特征子集.将该算法扩展到基于邻域粗糙集的实值和混合型不完整决策
系统,得到基于邻域粗糙集的不完整决策系统前向顺序特征选择算法.同时,将基于相容关系的不完整决策系统快速属性约简算法推]’一到实值和混合属性的不完整决策系统,得到适用于实值、混合属性的不
完整决策系统后向特征选择算法.理论分析和University of California lrvinc机器学习数据库数据集的实验共同表明,木文提出的基于邻域粗糙集的不完整决策系统前向特征选择算法有效降低了不完整决
策系统特征选择算法的时间复杂度,在保持系统识别能力的情况卜,用更少的时间得到决策系统的属性约简子集,即特征子集.然而,木文前向特征选择算法的缺陷是有可能因为无法选择到第一个最重要的
特征(属性)而使特征选择过程不能进行卜去,从而不能完成特征选择过程.
[1]Pawlak Z. Rough sets, international Journal of Computer and information Sciences, 1982,11 (5):341一356. [2]Pawlak Z. Rough sets; Theoretical aspects of rcasoning about data. Dordrccht:Kluwcr Aca dcmic publishers, 1991. [3]Kryszkicwicz M. Rough sct approach to mcom plete information systems, information Sci ences 1998,112(1一1):39一49. [4]Chmiclcwski M R,Urzymala-Bussc J W, Peterson N W , et al.The rule induction system LE RS一a vcrsion for personal computer. Founda tions of Computing and Decision Science, 1993 18(3一4)181一212. [5]Wang U Y. Extension of Rough set under in- complete information system. Journal of Com- puter Research and Development,2002,39 (10); 1238-1243.(王国胤. Rough集理论在不完备信息系统中的扩充.计算机研究与发展,2002,39(10):1238一1243}. [6]Meng Z Q, Shi Z Z. A fast approach to attrib ute reduction in incomplete decision systems witn tolerance relation based rough sets, lnfolr mation Sciences, 2009,179(16):2771一2793. [7]Hu Q H,Yu D R,Xie Z X. Neighhorhood classifiers. Expert Systems with Applications. 2008.34(2).866-876 [8]Hu Q H,Liu J F, Yu D R. Mixed feature seIcction based on granulation and approximation. Knowledge-Based System. 2008,21(4): 294一304. [9]Hu Q H,Yu D R,Liu J F, et al. Neighborhood rough set based heterogencous feature subset selection, Information Sciences, 2008,178 (18):3577一3591. [10]Wang H. Nearest neighbors by neighborhood counting. IEEE Transactions on Pattern Analy sis and Machine Intelligence 2006,28(6):942~953 [11]FranK A,Asuncion A. UCI Machine learning rcposnory [http;//archive. ics. uci. edu/ml]. lrvinc, CA:University of California, School of information and Computer Science, 2010. |
No related articles found! |
|