基于代价敏感的序贯三支决策最优粒度选择方法

张清华; 庞国弘; 李新太; 张雪秋

doi:10.11999/JEIT200821

基于代价敏感的序贯三支决策最优粒度选择方法

doi: 10.11999/JEIT200821

重庆邮电大学计算智能重庆市重点实验室重庆 400065

基金项目: 国家重点研发计划(2020YFC2003502)，国家自然科学基金(61876201)

详细信息

作者简介:
张清华：男，1974年生，教授，博士，博士生导师，研究方向为不确定信息处理、粗糙集与粒计算等

庞国弘：男，1994年生，硕士生，研究方向为不确定信息处理、粗糙集与三支决策

李新太：男，1995年生，硕士生，研究方向为不确定信息处理、数据挖掘与机器学习

张雪秋：女，1993年生，硕士生，研究方向为不确定信息处理、粗糙集与多尺度

通讯作者:
张清华　zhangqh@cqupt.edu.cn

中图分类号: TP301.6
计量
- 文章访问数: 1143
- HTML全文浏览量: 377
- PDF下载量: 55
- 被引次数: 0
出版历程
- 收稿日期: 2020-09-21
- 修回日期: 2021-07-19
- 网络出版日期: 2021-08-18
- 刊出日期: 2021-10-18

Optimal Granularity Selection Method Based on Cost-sensitive Sequential Three-way Decisions

Chongqing Key Laboratory of Computational Intelligence, Chongqing University of Posts and Telecommunications, Chongqing 400065, China

Funds: The National Key Research and Development Program of China (2020YFC2003502), The National Natural Science Foundation of China (61876201)

摘要

摘要: 最优粒度选择是序贯三支决策领域研究的热点之一，旨在通过合理的粒度选择来对复杂问题进行求解。在现阶段最优粒度选择中，代价敏感是影响决策的重要因素之一。针对这个问题，该文首先基于信息增益和卡方检验提出一种新的属性重要度计算方法；其次，为了更好地符合实际应用场景，在构建多粒度空间时将代价参数与粒度大小相结合，设置了相应的惩罚规则，并分析了决策阈值的变化规律；最后，为了消除测试代价和决策代价量纲不一致所带来的影响，借助变异系数设计了一种客观的代价计算方法。实验结果表明，该模型适用于现有代价认知场景，能在给定代价情况下选出代价最小的最优粒层。
- 序贯三支决策 /
- 属性重要度 /
- 惩罚函数 /
- 变异系数 /
- 最优粒度选择
Abstract: Optimal granularity selection is one of the hotspots in the research of sequential three-way decisions. It aims to solve complex problems through reasonable granularity selection. At present, in the field of optimal granularity selection, cost sensitivity is one of the important factors affecting decision making. To solve this problem, firstly, based on information gain and chi-squared test, a novel method to measure the attribute significance is proposed when constructing the multi-granularity space in this paper. Then, to better conform the practical application, the corresponding penalty rule is set by combining the cost parameters and the granularity, and the variation rule of the decision threshold is analyzed. Finally, to eliminate the influence of the dimensional difference between the test cost and the decision cost, an objective cost calculation method is given by the coefficient of variation. The experimental results show that the proposed algorithm can be used in existing cost cognition scene, and the optimal granular layer with the lowest cost can be obtained under the given cost scene.
- Sequential three-way decisions /
- Attribute significance /
- Penalty function /
- Coefficient of variation /
- Optimal granularity selection

HTML全文

图 1 多粒度空间的构造过程

下载: 全尺寸图片幻灯片

图 2 算法框架

下载: 全尺寸图片幻灯片

图 3 第1组代价参数下各数据集最优粒层的代价变化

下载: 全尺寸图片幻灯片

图 4 第2组代价参数下各数据集最优粒层的代价变化

下载: 全尺寸图片幻灯片

表 1 代价参数矩阵

	$X$	${}^\neg X$
${a_P}$	0	$\lambda _{{\rm{PN}}}^k$
${a_B}$	$\lambda _{{\rm{BP}}}^k$	$\lambda _{{\rm{BN}}}^k$
${a_N}$	$\lambda _{{\rm{NP}}}^k$	0

下载: 导出CSV

表 2 数据集的描述

序号	数据集	属性特征	数目	条件属性个数
1	Balance-scale	Categorical	625	4
2	Breast Cancer Wisconsin	Integer	699	9
3	Tic-Tac-Toe Endgame	Categorical	958	9
4	Car Evaluation	Categorical	1728	6
5	Nursery	Categorical	12960	8
6	Chess	Categorical, Integer	28056	6

下载: 导出CSV

表 3 代价参数

	${\lambda _{{\rm{PP}}}}$	${\lambda _{{\rm{BP}}}}$	${\lambda _{{\rm{NP}}}}$	${\lambda _{{\rm{PN}}}}$	${\lambda _{{\rm{BN}}}}$	${\lambda _{{\rm{NN}}}}$
第1组	0	1	4	5	2	0
第2组	0	2	6	7	3	0

下载: 导出CSV

表 4 第1组代价参数下各个数据集最优粒层信息

数据集	最优粒层	测试代价	决策代价	权重	总代价
Balance-scale	3	696.6	118.0	(0.46,0.54)	0.4172
Breast Cancer Wisconsin	8	323.7	522.6	(0.52,0.48)	0.3684
Tic-Tac-Toe Endgame	6	424.0	314.3	(0.49,0.51)	0.4453
Car Evaluation	4	636.0	35.2	(0.52,0.48)	0.4032
Chess	4	1997.3	0.0	(0.50,0.50)	0.4818
Nursery	6	998.9	8677.1	(0.54,0.46)	0.5423

下载: 导出CSV

表 5 第2组代价参数下每个数据集最优粒层信息

数据集	最优粒层	测试代价	决策代价	权重	总代价
Balance-scale	3	696.6	73.0	(0.47,0.53)	0.3172
Breast Cancer Wisconsin	5	227.9	652.1	(0.48,0.52)	0.4459
Tic-Tac-Toe Endgame	6	424.0	147.4	(0.40,0.60)	0.2941
Car Evaluation	4	636.0	372.1	(0.42,0.58)	0.3132
Chess	4	1997.3	14029.1	(0.41,0.59)	0.3705
Nursery	5	731.3	17162.085	(0.55,0.45)	0.5522

下载: 导出CSV

表 6 最优粒层比较

数据集	模型	最优粒层	冗余属性	最优属性子集
Balance-scale	模型1	3	$\phi $	$\{ {c_4},{c_3},{c_2}\} $
Balance-scale	模型2	3	$\phi $	$\{ {c_4},{c_3},{c_2}\} $
Breast Cancer Wisconsin	模型1	8	$\{ {c_1}\} $	$\{ {c_2},{c_3},{c_6},{c_7},{c_5},{c_8},{c_4},{c_9}\} $
Breast Cancer Wisconsin	模型2	5	$\{ {c_1}\} $	$\{ {c_2},{c_3},{c_6},{c_7},{c_5}\} $
Tic-Tac-Toe Endgame	模型1	6	$\phi $	$\{ {c_8},{c_6},{c_4},{c_2},{c_9},{c_7}\} $
Tic-Tac-Toe Endgame	模型2	6	$\phi $	$\{ {c_8},{c_6},{c_4},{c_2},{c_9},{c_7}\} $
Car Evaluation	模型1	3	$\{ {c_2}\} $	$\{ {c_4},{c_1},{c_6}\} $
Car Evaluation	模型2	3	$\{ {c_2}\} $	$\{ {c_4},{c_1},{c_6}\} $
Chess	模型1	6	$\phi $	$\{ {c_3},{c_5},{c_2},{c_1},{c_4},{c_6}\} $
Chess	模型2	5	$\phi $	$\{ {c_3},{c_5},{c_2},{c_1},{c_4}\} $
Nursery	模型1	6	$\{ {c_3},{c_6}\} $	$\{ {c_8},{c_2},{c_1},{c_7},{c_5},{c_4}\} $
Nursery	模型2	6	$\{ {c_3},{c_6}\} $	$\{ {c_8},{c_2},{c_1},{c_7},{c_5},{c_4}\} $

下载: 导出CSV

参考文献(24)

[1]	LIAO Shujiao, ZHU Qingxin, QIAN Yuhua, et al. Multi-granularity feature selection on cost-sensitive data with measurement errors and variable costs[J]. Knowledge-Based Systems, 2018, 158: 25–42. doi: 10.1016/j.knosys.2018.05.020
[2]	YANG Jie, WANG Guoyin, ZHANG Qinghua, et al. Optimal granularity selection based on cost-sensitive sequential three-way decisions with rough fuzzy sets[J]. Knowledge-Based Systems, 2019, 163: 131–144. doi: 10.1016/j.knosys.2018.08.019
[3]	LI Huaxiong, ZHANG Libo, ZHOU Xianzhong, et al. Cost-sensitive sequential three-way decision modeling using a deep neural network[J]. International Journal of Approximate Reasoning, 2017, 85: 68–78. doi: 10.1016/j.ijar.2017.03.008
[4]	ZHANG Yanping, ZOU Huijin, CHEN Xi, et al. Cost-sensitive three-way decisions model based on CCA[C]. The 9th International Conference on Rough Sets and Current Trends in Computing, Granada and Madrid, Spain, 2014: 172–180. doi: 10.1007/978-3-319-08644-6_18.
[5]	JIA Xiuyi, LIAO Wenhe, TANG Zhenmin, et al. Minimum cost attribute reduction in decision-theoretic rough set models[J]. Information Sciences, 2013, 219: 151–167. doi: 10.1016/j.ins.2012.07.010
[6]	YANG Xibei, QI Yunsong, SONG Xiaoning, et al. Test cost sensitive multigranulation rough set: Model and minimal cost selection[J]. Information Sciences, 2013, 250: 184–199. doi: 10.1016/j.ins.2013.06.057
[7]	MIN Fan and LIU Qihe. A hierarchical model for test-cost-sensitive decision systems[J]. Information Sciences, 2009, 179(14): 2442–2452. doi: 10.1016/j.ins.2009.03.007
[8]	JU Hengrong, LI Huaxiong, YANG Xibei, et al. Cost-sensitive rough set: A multi-granulation approach[J]. Knowledge-Based Systems, 2017, 123: 137–153. doi: 10.1016/j.knosys.2017.02.019
[9]	JU Hengrong, YANG Xibei, YU Hualong, et al. Cost-sensitive rough set approach[J]. Information Sciences, 2016, 355/356: 282–298. doi: 10.1016/j.ins.2016.01.103
[10]	YAO Yiyu and DENG Xiaofei. Sequential three-way decisions with probabilistic rough sets[C]. IEEE 10th International Conference on Cognitive Informatics and Cognitive Computing (ICCI-CC’11), Banff, Canada, 2011: 120–125. doi: 10.1109/COGINF.2011.6016129.
[11]	ZHANG Qinghua, CHEN Yuhong, YANG Jie, et al. Fuzzy entropy: A more comprehensible perspective for interval shadowed sets of fuzzy sets[J]. IEEE Transactions on Fuzzy Systems, 2020, 28(11): 3008–3022. doi: 10.1109/tfuzz.2019.2947224
[12]	ZHANG Qinghua, ZHAO Fan, YANG Jie, et al. Three-way decisions of rough vague sets from the perspective of fuzziness[J]. Information Sciences, 2020, 523: 111–132. doi: 10.1016/j.ins.2020.03.013
[13]	张清华, 幸禹可, 周玉兰. 基于粒计算的增量式知识获取方法[J]. 电子与信息学报, 2011, 33(2): 435–441. doi: 10.3724/SP.J.1146.2010.00217 ZHANG Qinghua, XING Yuke, and ZHOU Yulan. The incremental knowledge acquisition algorithm based on granular computing[J]. Journal of Electronics &Information Technology, 2011, 33(2): 435–441. doi: 10.3724/SP.J.1146.2010.00217
[14]	FANG Yu, GAO Cong, and YAO Yiyu. Granularity-driven sequential three-way decisions: A cost-sensitive approach to classification[J]. Information Sciences, 2020, 507: 644–664. doi: 10.1016/j.ins.2019.06.003
[15]	LI Huaxiong, ZHANG Libo, HUANG Bing, et al. Sequential three-way decision and granulation for cost-sensitive face recognition[J]. Knowledge-Based Systems, 2016, 91: 241–251. doi: 10.1016/j.knosys.2015.07.040
[16]	QIAN Jin, LIU Caihui, MIAO Duoqian, et al. Sequential three-way decisions via multi-granularity[J]. Information Sciences, 2020, 507: 606–629. doi: 10.1016/j.ins.2019.03.052
[17]	陈泽华, 马贺. 基于粒矩阵的多输入多输出真值表快速并行约简算法[J]. 电子与信息学报, 2015, 37(5): 1260–1265. doi: 10.11999/JEIT141129 CHEN Zehua and MA He. Granular matrix based rapid parallel reduction algorithm for MIMO truth table[J]. Journal of Electronics &Information Technology, 2015, 37(5): 1260–1265. doi: 10.11999/JEIT141129
[18]	ZHANG Qinghua, XIA Deyou, and WANG Guoyin. Three-way decision model with two types of classification errors[J]. Information Sciences, 2017, 420: 431–453. doi: 10.1016/j.ins.2017.08.066
[19]	王国胤. Rough集理论与知识获取[M]. 西安: 西安交通大学出版社, 2001: 17–18.
[20]	PAWLAK Z. Rough sets[J]. International Journal of Computer & Information Sciences, 1982, 11(5): 341–356. doi: 10.1007/BF01001956
[21]	PLACKETT R L. Karl Pearson and the chi-squared test[J]. International Statistical Review, 1983, 51(1): 59–72. doi: 10.2307/1402731
[22]	周志华. 机器学习[M]. 北京: 清华大学出版社, 2016: 75.
[23]	QUINLAN J R. Induction of decision trees[J]. Machine Learning, 1986, 1(1): 81–106. doi: 10.1023/A:1022643204877
[24]	LIU Dun, LI Tianrui, and LIANG Decui. Three-way decisions in dynamic decision-theoretic rough sets[C]. The 8th International Conference on Rough Sets and Knowledge Technology, Halifax, Canada, 2013: 291–301. doi: 10.1007/978-3-642-41299-8_28.