正在加载图片...
第6期 刘牧雷,等:基于三支决策的序列数据代价敏感分类算法 ·1261· 型建立可信度判据。结合三支决策理论,在时间 risk prediction using cost-sensitive with Nonnegativity- 序列分析问题中,三支决策模型可以为预测结果 constrained Autoencoders based on imbalanced naturalist- 增加可信度的判据,使得预测结果更加具有分析 ic driving data[J/OL].IEEE transactions on intelligent 和处理的价值。 transportation systems:(2019-01-17).https://ieeexplore. 但是当前的工作只是初步的验证有关于深度 ieee.org/document/8617709.DOI:10.1109/TITS.2018. 2886280. 学习与三支决策相结合形成新的代价敏感分类的 [8]YAO Yiyu.Three-way decision:an interpretation of rules 初步研究。本文的研究尚处于初步的阶段。未 in rough set theory[C]//Proceedings of the 4th Internation- 来,对于模型的改进仍有许多研究空间。例如, al Conference on Rough Sets and Knowledge Technology. 对于三支决策算法,可以结合新的边界理论,形 Gold Coast,Australia,2009:642-649. 成自动化的边界确定;在整体模型中,可以借助 [9]GERS F A,SCHMIDHUBER J,CUMMINS F.Learning boost或专家分类器等模型,提出更完善的理论; to forget:continual prediction with LSTM[J].Neural com- 以及结合Alex-net等其他更高效的分类器来进一 putation,2000,12(10):2451-2471. 步提高前置分类器的性能等。这些改进都将能够 [10]ELKAN C.The foundations of cost-sensitive learning 进一步提高三支决策在在代价敏感分类领域的应 [C]//Proceedings of the 17th International Joint Confer- 用频率。 ence of Artificial Intelligence.Morgan Kaufmann,Seattle, 2001:973-978 参考文献: [11]LIU Xuying,ZHOU Zhihua.The influence of class im- [1]KARIM F,MAJUMDAR S,DARABI H,et al.LSTM balance on cost-sensitive learning:an empirical study [Cl//Proceedings of the 6th International Conference on fully convolutional networks for time series classification Data Mining.Hong Kong,China,2006:970-974. [J.IEEE access,.2018,6:1662-1669 [12]ZADROZNY B,LANGFORD J,ABE N.Cost-sensitive [2]KARIM F.MAJUMDAR S.DARABI H.et al.Multivari- learning by cost-proportionate example weighting ate LSTM-FCNs for time series classification[J].Neural [C]//Proceedings of the 3rd IEEE International Confer- networks,2019,116:237-245. [3]KHAN S H,HAYAT M,BENNAMOUN M,et al.Cost- ence on Data Mining.Melbourne,FL,USA,2003: 435-442 sensitive learning of deep feature representations from im- balanced data[J].IEEE transactions on neural networks and 作者简介: learning systems,2018,29(8):3573-3587. 刘牧雷,男,1993年生,硕士研究 [4]FERNANDEZ A,GARCiA S,GALAR M,et al.Cost- 生,主要研究方向为三支决策、代价敏 sensitive learning[M]//FERNANDEZ A,GARCIA S. 感分类。 GALAR M,et al.Learning from Imbalanced Data Sets. Cham:Springer,2018:63-78 [5]YAN Ke,MA Lulu,DAI Yuting,et al.Cost-sensitive and sequential feature selection for chiller fault detection and diagnosis[J].International journal of refrigeration,2018, 徐菲菲,女,1983年生,副教授 86:401-409 中国计算机学会和中国人工智能学会 会员,主要研究方向为粒计算理论、粗 [6]JIANG Xinxin,PAN Shirui,LONG Guodong,et al.Cost- 糙集理论、数据挖掘、人工智能与机器 sensitive parallel learning framework for insurance intelli- 学习。主持国家自然科学基金项目 gence operation[J].IEEE transactions on industrial elec- 1项:上海市教育发展基金会和上海 tronics,2019,66(12):9713-9723 市教育委员会“晨光计划”1项、上海市 [7]CHEN Jie,WU Zhongcheng,ZHANG Jun.Driving safety 教育委员会科研创新项目1项等。型建立可信度判据。结合三支决策理论,在时间 序列分析问题中,三支决策模型可以为预测结果 增加可信度的判据,使得预测结果更加具有分析 和处理的价值。 但是当前的工作只是初步的验证有关于深度 学习与三支决策相结合形成新的代价敏感分类的 初步研究。本文的研究尚处于初步的阶段。未 来,对于模型的改进仍有许多研究空间。例如, 对于三支决策算法,可以结合新的边界理论,形 成自动化的边界确定;在整体模型中,可以借助 boost 或专家分类器等模型,提出更完善的理论; 以及结合 Alex-net 等其他更高效的分类器来进一 步提高前置分类器的性能等。这些改进都将能够 进一步提高三支决策在在代价敏感分类领域的应 用频率。 参考文献: KARIM F, MAJUMDAR S, DARABI H, et al. LSTM fully convolutional networks for time series classification [J]. IEEE access, 2018, 6: 1662–1669. [1] KARIM F, MAJUMDAR S, DARABI H, et al. Multivari￾ate LSTM-FCNs for time series classification[J]. Neural networks, 2019, 116: 237–245. [2] KHAN S H, HAYAT M, BENNAMOUN M, et al. Cost￾sensitive learning of deep feature representations from im￾balanced data[J]. IEEE transactions on neural networks and learning systems, 2018, 29(8): 3573–3587. [3] FERNÁNDEZ A, GARCÍA S, GALAR M, et al. Cost￾sensitive learning[M]//FERNÁNDEZ A, GARCÍA S, GALAR M, et al. Learning from Imbalanced Data Sets. Cham: Springer, 2018: 63-78. [4] YAN Ke, MA Lulu, DAI Yuting, et al. Cost-sensitive and sequential feature selection for chiller fault detection and diagnosis[J]. International journal of refrigeration, 2018, 86: 401–409. [5] JIANG Xinxin, PAN Shirui, LONG Guodong, et al. Cost￾sensitive parallel learning framework for insurance intelli￾gence operation[J]. IEEE transactions on industrial elec￾tronics, 2019, 66(12): 9713–9723. [6] [7] CHEN Jie, WU Zhongcheng, ZHANG Jun. Driving safety risk prediction using cost-sensitive with Nonnegativity￾constrained Autoencoders based on imbalanced naturalist￾ic driving data[J/OL]. IEEE transactions on intelligent transportation systems: (2019-01-17). https://ieeexplore. ieee.org/document/8617709. DOI: 10.1109/TITS.2018. 2886280. YAO Yiyu. Three-way decision: an interpretation of rules in rough set theory[C]//Proceedings of the 4th Internation￾al Conference on Rough Sets and Knowledge Technology. Gold Coast, Australia, 2009: 642–649. [8] GERS F A, SCHMIDHUBER J, CUMMINS F. Learning to forget: continual prediction with LSTM[J]. Neural com￾putation, 2000, 12(10): 2451–2471. [9] ELKAN C. The foundations of cost-sensitive learning [C]//Proceedings of the 17th International Joint Confer￾ence of Artificial Intelligence. Morgan Kaufmann, Seattle, 2001: 973–978. [10] LIU Xuying, ZHOU Zhihua. The influence of class im￾balance on cost-sensitive learning: an empirical study [C]//Proceedings of the 6th International Conference on Data Mining. Hong Kong, China, 2006: 970–974. [11] ZADROZNY B, LANGFORD J, ABE N. Cost-sensitive learning by cost-proportionate example weighting [C]//Proceedings of the 3rd IEEE International Confer￾ence on Data Mining. Melbourne, FL, USA, 2003: 435–442. [12] 作者简介: 刘牧雷,男,1993 年生,硕士研究 生,主要研究方向为三支决策、代价敏 感分类。 徐菲菲,女,1983 年生,副教授, 中国计算机学会和中国人工智能学会 会员,主要研究方向为粒计算理论、粗 糙集理论、数据挖掘、人工智能与机器 学习。主持国家自然科学基金项目 1 项;上海市教育发展基金会和上海 市教育委员会“晨光计划”1 项、上海市 教育委员会科研创新项目 1 项等。 第 6 期 刘牧雷,等:基于三支决策的序列数据代价敏感分类算法 ·1261·
<<向上翻页
©2008-现在 cucdc.com 高等教育资讯网 版权所有