正在加载图片...
第6期 程龙,等:弱标记不完备决策系统的增量式属性约简算法 ·1089· 和Letter Recognition数据集中两者的属性约简结 性能较优的属性约简结果。为大规模复杂数据的 果完全相同。由此可知,本文提出的WIDIAR算 属性约简问题,提供了一个可行的增量式属性约 法能够有效节约大量计算时间,同时能获取分类 简方法。 表4属性约简结果对比 Table 4 Comparison of attribute reduction results 数据集 WIDIAR算法 WIDAR算法 Automobile 13,25,2 24,7,6,25,1 Soybean(Large) 17.6.15.10,17.435,12.8321,30 1.7.6.15,1035.4213.22,2.289.13,8.17.1116.19.24,30.5,12.14,31 Dermatology 34.4,32,3,16,1,2 4,319.2,32,1734 Cylinder Bands 23525,3 2,29,22,3 Mushroom 9,3,22,1,2,15,5,21,14,13,20,12,17,7,6 9,3,22,1,2,15,5,21,14,13,20,12,7,17,6 Letter Recognition 2,15,8,9,11,3,6,4,7,1,12,10,5,16,13,14 2.1589,11,3,6,47,1.1210,5,16.13.14 6结束语 WANG Yinglong,ZENG Qi,QIAN Wenbin,et al.Incre- mental attribute reduction method for incomplete hybrid 在众多的现实应用领域中,由于数据采集技 data with variable precision[J.Journal of computer applic- 术和成本的限制,存在大量的不完备高维数据, ations,.2018.38(10):2764-2771 且通常只有少量的数据存在标记信息。若仅利 [5]MA Fumin,DING Mianwei,ZHANG Tengfei,et al.Com- 用有标记的数据获取属性约简结果,将会丢失部 pressed binary discernibility matrix based incremental at- 分有效的信息。为了有效地利用弱标记不完备 tribute reduction algorithm for group dynamic data[J]. 的高维数据,从中获取分类效果更优的属性约简 Neurocomputing,2019,344:20-27. 集,本文基于实例的区分对,提出属性相对重要 [6]QIAN Yuhua,LIANG Jiye,PEDRYCZ W,et al.Positive 度的概念,设计了启发式属性约简算法。针对动 approximation:an accelerator for attribute reduction in 态变化场景,详细分析了数据的动态变化对属性 rough set theory[].Artificial intelligence,2010,174(9/10): 约简的影响和更新机制。并在此基础上,提出了 597-618. 增量式的属性约简算法。实验结果表明,该算法 [7]LIANG Jiye,MI Junrong,WEI Wei,et al.An accelerator 在处理大规模数据时,相比静态属性约简算法, for attribute reduction based on perspective of objects and attributes[J].Knowledge-based systems,2013.44:90-100. 在保证分类性能的同时,能够高效地获取属性约 [8]DAI Jianhua,HU Qinghua,HUHu,et al.Neighbor incon- 简结果。下一步工作将拓展增量式属性约简算 sistent pair selection for attribute reduction by rough set 法应用场景,考虑属性集变化后,属性约简集的 approach[J].IEEE transactions on fuzzy systems,2018, 更新问题。 26(2):937-950. 参考文献: [9]TENG Shuhua,LU Min,YANG AFeng,et al.Efficient at- tribute reduction from the viewpoint of discernibility[J. [1]PAWLAK Z,SKOWRON A.Rough sets:some exten- Information sciences,2016,326:297-314 sions[J].Information sciences,2007,177(1):28-40. [10]QIAN Yuhua,LIANG Jiye,LI Deyu,et al.Approxima- [2]王国胤,姚一豫,于洪.粗糙集理论与应用研究综述) tion reduction in inconsistent incomplete decision 计算机学报,2009,32(7):1229-1246 tables[J].Knowledge-based systems,2010,23(5): WANG Guoyin,YAO Yiyu,YU Hong.A survey on rough 427-433. set theory and applications[J].Chinese journal of com- [11]MENG Zugiang,SHI Zhongzhi.A fast approach to attrib- puters,2009,32(7):1229-1246. ute reduction in incomplete decision systems with toler- [3]HU Qinghua,LIU Jinfu,YU Daren.Mixed feature selec- ance relation-based rough sets[J].Information sciences, tion based on granulation and approximation[J].Know- 2009,179(16):27742793. ledge-based systems,2008,21(4):294-304. [12]XIE Xiaojun,QIN Xiaolin.A novel incremental attribute [4]王映龙,曾淇,钱文彬,等.变精度下不完备混合数据的 reduction approach for dynamic incomplete decision sys- 增量式属性约简方法[].计算机应用,2018,38(10): tems[J].International journal of approximate reasoning 2764-2771. 2018,93:443-462和 Letter Recognition 数据集中两者的属性约简结 果完全相同。由此可知,本文提出的 WIDIAR 算 法能够有效节约大量计算时间,同时能获取分类 性能较优的属性约简结果。为大规模复杂数据的 属性约简问题,提供了一个可行的增量式属性约 简方法。 表 4 属性约简结果对比 Table 4 Comparison of attribute reduction results 数据集 WIDIAR算法 WIDAR算法 Automobile 13,25,2 24,7,6,25,1 Soybean(Large) 1,7,6,15,10,17,4,35,12,8,3,21,30 1,7,6,15,10,35,4,21,3,22,2,28,9,13,8,17,11,16,19,24,30,5,12,14,31 Dermatology 34,4,32,3,16,1,2 4,3,19,2,32,17,34 Cylinder Bands 2,35,25,3 2,29,22,3 Mushroom 9,3,22,1,2,15,5,21,14,13,20,12,17,7,6 9,3,22,1,2,15,5,21,14,13,20,12,7,17,6 Letter Recognition 2,15,8,9,11,3,6,4,7,1,12,10,5,16,13,14 2,15,8,9,11,3,6,4,7,1,12,10,5,16,13,14 6 结束语 在众多的现实应用领域中,由于数据采集技 术和成本的限制,存在大量的不完备高维数据, 且通常只有少量的数据存在标记信息。若仅利 用有标记的数据获取属性约简结果,将会丢失部 分有效的信息。为了有效地利用弱标记不完备 的高维数据,从中获取分类效果更优的属性约简 集,本文基于实例的区分对,提出属性相对重要 度的概念,设计了启发式属性约简算法。针对动 态变化场景,详细分析了数据的动态变化对属性 约简的影响和更新机制。并在此基础上,提出了 增量式的属性约简算法。实验结果表明,该算法 在处理大规模数据时,相比静态属性约简算法, 在保证分类性能的同时,能够高效地获取属性约 简结果。下一步工作将拓展增量式属性约简算 法应用场景,考虑属性集变化后,属性约简集的 更新问题。 参考文献: PAWLAK Z, SKOWRON A. Rough sets: some exten￾sions[J]. Information sciences, 2007, 177(1): 28–40. [1] 王国胤, 姚一豫, 于洪. 粗糙集理论与应用研究综述 [J]. 计算机学报, 2009, 32(7): 1229–1246. WANG Guoyin, YAO Yiyu, YU Hong. A survey on rough set theory and applications[J]. Chinese journal of com￾puters, 2009, 32(7): 1229–1246. [2] HU Qinghua, LIU Jinfu, YU Daren. Mixed feature selec￾tion based on granulation and approximation[J]. Know￾ledge-based systems, 2008, 21(4): 294–304. [3] 王映龙, 曾淇, 钱文彬, 等. 变精度下不完备混合数据的 增量式属性约简方法 [J]. 计算机应用, 2018, 38(10): 2764–2771. [4] WANG Yinglong, ZENG Qi, QIAN Wenbin, et al. Incre￾mental attribute reduction method for incomplete hybrid data with variable precision[J]. Journal of computer applic￾ations, 2018, 38(10): 2764–2771. MA Fumin, DING Mianwei, ZHANG Tengfei, et al. Com￾pressed binary discernibility matrix based incremental at￾tribute reduction algorithm for group dynamic data[J]. Neurocomputing, 2019, 344: 20–27. [5] QIAN Yuhua, LIANG Jiye, PEDRYCZ W, et al. Positive approximation: an accelerator for attribute reduction in rough set theory[J]. Artificial intelligence, 2010, 174(9/10): 597–618. [6] LIANG Jiye, MI Junrong, WEI Wei, et al. An accelerator for attribute reduction based on perspective of objects and attributes[J]. Knowledge-based systems, 2013, 44: 90–100. [7] DAI Jianhua, HU Qinghua, HU Hu, et al. Neighbor incon￾sistent pair selection for attribute reduction by rough set approach[J]. IEEE transactions on fuzzy systems, 2018, 26(2): 937–950. [8] TENG Shuhua, LU Min, YANG AFeng, et al. Efficient at￾tribute reduction from the viewpoint of discernibility[J]. Information sciences, 2016, 326: 297–314. [9] QIAN Yuhua, LIANG Jiye, LI Deyu, et al. Approxima￾tion reduction in inconsistent incomplete decision tables[J]. Knowledge-based systems, 2010, 23(5): 427–433. [10] MENG Zuqiang, SHI Zhongzhi. A fast approach to attrib￾ute reduction in incomplete decision systems with toler￾ance relation-based rough sets[J]. Information sciences, 2009, 179(16): 2774–2793. [11] XIE Xiaojun, QIN Xiaolin. A novel incremental attribute reduction approach for dynamic incomplete decision sys￾tems[J]. International journal of approximate reasoning, 2018, 93: 443–462. [12] 第 6 期 程龙,等:弱标记不完备决策系统的增量式属性约简算法 ·1089·
<<向上翻页向下翻页>>
©2008-现在 cucdc.com 高等教育资讯网 版权所有