Study on the Approximate Reasoning Models of Decision Implication in Formal Decision Context
-
摘要: 基于形式背景获取决策蕴涵、概念规则等知识是数据分析、机器学习的重要研究内容之一. 首先, 利用属性逻辑语义对决策蕴涵的特性进行刻画. 其次, 在经典二值逻辑框架下分析决策蕴涵、概念规则的基于全蕴涵三I推理思想及分离规则(Modus ponens, MP)和逆分离规则(Modus tonens, MT)的近似推理模式的特征, 证明决策蕴涵的MP、MT近似推理结论是决策蕴涵, 概念规则的MP、MT近似推理结论是概念规则等结论. 引进属性逻辑公式的伪距离, 在属性逻辑伪距离空间中分析推理对象范围参数变化对决策蕴涵MP、MT近似推理结论的影响. 最后, 提出若干通过MP、MT近似推理生成决策蕴涵、概念规则及拟决策蕴涵的模式和方法, 数值实验验证了所提方法的有效性.Abstract: It is one of the contents of data analysis and machine learning to obtain the knowledge of decision implication and conceptual rule from formal decision context. Firstly, attribute logic semantics is used to analyze the characteristics of decision implication. Secondly, in the classical binary logic framework, the characteristics of modus ponens (MP) and modus tonens (MT) approximate reasoning models based on the all implication three-I inference idea is analyzed. It is proved that the MP and MT approximate reasoning conclusions of decision implications are decisions implications, the MP and MT approximate reasoning conclusions of concept rules are concept rules. The pseudo-metric of attribute logic formula is introduced, and the influence of MP and MT approximate reasoning conclusions of decision implication as the reasoning object range parameter changes in attribute logic pseudo-metric space is analyzed. Finally, some models and methods of generating decision implication, concept rules and quasi-decision implication through MP and MT approximate reasoning based on existing decision implication, concept rules and quasi-decision implication are proposed, numerical experiments show that the proposed methods are effective.
-
表 1 形式背景$K=(G,M,I)$
Table 1 A formal context $K=(G,M,I)$
$G$ $a_1$ $a_2$ $a_3$ $a_4$ $a_5$ $u_1$ 1 1 0 0 1 $u_2$ 0 0 1 1 0 $u_3$ 1 0 1 0 0 $u_4$ 0 1 0 1 0 $u_5$ 0 1 0 1 1 $u_6$ 1 0 1 0 1 表 2 决策形式背景$K=(G,C,D,I,J)$
Table 2 A formal decision context $K=(G,C,D,I,J)$
$G$ $a_1$ $a_2$ $a_3$ $a_4$ $d_1$ $d_2$ $d_3$ $u_1$ 1 1 1 1 1 0 1 $u_2$ 0 0 1 0 0 1 0 $u_3$ 1 0 0 0 0 1 1 $u_4$ 0 1 1 0 1 0 0 表 3 决策形式背景$K=(G,C,D,I,J)$
Table 3 A formal decision context $K=(G,C,D,I,J)$
$G$ $a_1$ $a_2$ $a_3$ $a_4$ $a_5$ $a_6$ $d_1$ $d_2$ $d_3$ $u_1$ 1 1 0 0 1 1 1 0 1 $u_2$ 0 0 1 1 1 1 1 1 0 $u_3$ 1 0 1 0 0 1 0 0 1 $u_4$ 0 1 0 1 0 1 1 0 0 $u_5$ 0 1 0 1 1 1 1 1 0 $u_6$ 1 0 1 0 1 1 0 1 1 $u_7$ 0 1 1 1 0 0 1 1 1 表 4 生成的拟决策蕴涵个数
Table 4 The number of generated quasi-decision implications
数据组别 生成的拟决策蕴涵个数 数据组1 62 数据组2 58 数据组3 74 数据组4 71 表 5 生成的拟决策蕴涵后件与后件合取公式的伪距离
Table 5 The pseudo-metric between the consequent of generated quasi-decision implications and the consequent conjunctive
数据组别 最大伪距离 最小伪距离 数据组1 0.195 0.147 数据组2 0.189 0.160 数据组3 0.197 0.152 数据组4 0.194 0.161 表 6 测试数据变化生成的拟决策蕴涵表
Table 6 A table of generated quasi-decision implication as test data changes
数据组别 拟决策蕴涵个数 最小伪距离 数据组1 3 0.0082 数据组2 5 0.0157 数据组3 8 0.0256 数据组4 12 0.0324 数据组5 16 0.0418 数据组6 21 0.0527 数据组7 25 0.0619 数据组8 29 0.0718 数据组9 34 0.0821 数据组10 39 0.0913 表 7 后件集对结论的支持度和获取拟决策蕴涵时间成本对比
Table 7 Comparison of the support degree of consequent set to the conclusion and time consumption of obtaining quasi-decision implications
数据组别 $L$的后件集对结论支持度$\tau_{\Delta}$ 方式1)总时间成本(s) 方式2)总时间成本 (s) 增量获取所用时间成本(s) 数据组1 0.933 0.204 0.223 0.103 数据组2 0.923 0.283 0.304 0.147 数据组3 0.914 0.361 0.389 0.198 数据组4 0.903 0.450 0.484 0.258 数据组5 0.893 0.539 0.258 0.326 -
[1] 张文修, 梁怡, 吴伟志. 信息系统与知识发现. 北京: 科学出版社, 2003.Zhang Wen-Xiu, Liang Yi, Wu Wei-Zhi. Information Systems and Knowledge Discoveries. Beijing: Science Press, 2003. [2] 李金海, 魏玲, 张卓, 翟岩慧, 张涛, 智慧来, 等. 概念格理论与方法及其研究展望. 模式识别与人工智能, 2020, 33(7): 619−642Li Jin-Hai, Wei Ling, Zhang Zhuo, Zhai Yan-Hui, Zhang Tao, Zhi Hui-Lai, et al. Concept lattice theory and method and their research prospect. Pattern Recognition and Artificial Intelligence, 2020, 33(7): 619−642 [3] Ganter B, Wille R. Formal Concept Analysis Mathematical Foundations 3. Berlin: Germany Springer-Verlag, 1999. [4] Li J H, Mei C L, Kumar C A, Zhang X. On rule acquisition in decision formal contexts. International Journal of Machine Learning and Cybernetics, 2013, 4(6): 721−731 doi: 10.1007/s13042-013-0150-z [5] Zhai Y H, Li D Y, Qu K S. Decision implications: A logical point of view. International Journal of Machine Learning and Cybernetics, 2014, 5: 509−516 doi: 10.1007/s13042-013-0204-2 [6] Zhai Y H, Li D Y, Qu K S. Fuzzy decision implications. Knowledge-Based Systems, 2013, 37: 230−236 doi: 10.1016/j.knosys.2012.08.004 [7] 贾楠, 张少霞, 翟岩慧, 李德玉. 决策蕴涵上的推理规则和推理过程研究. 计算机科学与探索, 2020, 14(2): 344−352Jia Nan, Zhang Shao-Xia, Zhai Yan-Hui, Li De-Yu. Study of inference rules and deduction process on decision implications. Journal of Frontiers of Computer Science and Technology, 2020, 14(2): 344−352 [8] 王琪, 李德玉, 翟岩慧, 张少霞. 含参决策蕴涵. 计算机研究与发展, 2022, 59(9): 2066−2074Wang Qi, Li De-Yu, Zhai Yan-Hui, Zhang Shao-Xia. Parameterized fuzzy decision implication. Journal of Computer Research and Development, 2022, 59(9): 2066−2074 [9] 张少霞, 李德玉, 翟岩慧. 逻辑型决策蕴涵. 郑州大学学报 (理学版), 2020, 52(1): 1−7Zhang Shao-Xia, Li De-Yu, Zhai Yan-Hui. Logic-type decision implication. Journal of Zhengzhou University (Natural Science), 2020, 52(1): 1−7 [10] Zhang S X, Li D Y, Zhai Y H, Kang X P. A comparative study of decision implication, concept rule and granule. Information Sciences, 2020, 508: 33−49 doi: 10.1016/j.ins.2019.08.053 [11] 李金海, 李玉斐, 米允龙, 吴伟志. 多粒度形式概念分析的介粒度标记方法. 计算机研究与发展, 2020, 57(2): 447−458Li Jin-Hai, Li Yu-Fei, Mi Yun-Long, Wu Wei-Zhi. Meso-granularity labeled method for multi-granularity formal concept analysis. Journal of Computer Research and Development, 2020, 57(2): 447−458 [12] 李金海, 王飞, 吴伟志, 徐伟华, 杨习贝, 折延宏. 基于粒计算的多粒度数据分析方法综述. 数据采集与处理, 2021, 36(3): 418−435Li Jin-Hai, Wang Fei, Wu Wei-Zhi, Xu Wei-Hua, Yang Xi-Bei, She Yan-Hong. Review of multi-granularity data analysis methods based on granular computing. Journal of Data Acquisition and Processing, 2021, 36(3): 418−435 [13] Wang G J, Zhou H J. Quantitative logic. Information Science, 2009, 179: 226−247 doi: 10.1016/j.ins.2008.09.008 [14] Linden G, Smith B, York J. Amazon.com recommendations: Item-to-item collaborative filtering. IEEE Internet Computing, 2003, 7(1): 76−80 doi: 10.1109/MIC.2003.1167344 [15] Freyer D A, Hsieh Y H, Levin S R, Pines J M, Mareiniss D P, Mohareb A, et al. Goolgle flu trends: Correlation with emergency department influenza rates and crowding metrics. Clinical Infectious Diseases, 2012, 54(4): 463−469 doi: 10.1093/cid/cir883 [16] 梁吉业, 冯晨娇, 宋鹏. 大数据相关分析综述. 计算机学报, 2016, 39(1): 1−18Liang Ji-Ye, Feng Chen-Jiao, Song Peng. A survey on correlation analysis of big data. Chinese Journal of Computers, 2016, 39(1): 1−18 [17] 张家录, 吴霞. 属性逻辑的近似推理模式研究. 计算机科学与探索, 2023, 17(4): 848−856Zhang Jia-Lu, Wu Xia. Study on the approximate reasoning models based on the attribute logic. Journal of Frontiers of Computer Science and Technology, 2023, 17(4): 848−856 [18] 王国俊. 非经典数理逻辑与近似推理. 北京: 科学出版社, 2000.Wang Guo-Jun. Non-classical Mathematical Logic and Approximate Reasoning. Beijing: Science Press, 2000. [19] 王国俊. 模糊推理的全蕴涵三I算法. 中国科学(E辑), 1999, 29(1): 43−53Wang Guo-Jun. The triple I algorithm for all implication of fuzzy reasoning. Science in China (Series E), 1999, 29(1): 43−53 [20] 张文修, 魏玲, 祁建军. 概念格的属性约简理论与方法. 中国科学(E辑), 2005, 35(6): 628−639Zhang Wen-Xiu, Wei Ling, Qi Jian-Jun. Attribute reduction theory and approach to concept lattice. Science in China (Series E), 2005, 35(6): 628−639 [21] 罗佳琪, 吴霞, 张家录, 杜佳甜. 基于形式概念的对象粒的属性公式描述. 模糊系统与数学, 2020, 34(1): 41−48Luo Jia-Qi, Wu Xia, Zhang Jia-Lu, Du Jia-Tian. The description of object granules by using attribute logic formulas based on formal concept. Fuzzy Systems and Mathematics, 2020, 34(1): 41−48 [22] Zhang J L, Wu X, Lu R H. Decision analysis methods combining quantitative logic and fuzzy soft sets. International Journal of Fuzzy Systems, 2020, 22(6): 1801−1814 doi: 10.1007/s40815-020-00899-6 [23] 吴霞, 张家录. 基于面向属性(对象)概念的对象粒的属性公式描述. 模式识别与人工智能, 2020, 33(9): 767−775Wu Xia, Zhang Jia-Lu. Attribute logical formulas description of object granules based on the property-oriented (object-oriented) concepts. Pattern Recognition and Artificial Intelligence, 2020, 33(9): 767−775 [24] 张家录, 陈雪刚, 赵晓东. 经典命题逻辑的概率语义及其应用. 计算机学报, 2014, 37(8): 1775−1785Zhang Jia-Lu, Chen Xue-Gang, Zhao Xiao-Dong. Theory of probability semantics of classical propositional logic and its application. Chinese Journal of Computers, 2014, 37(8): 1775−1785 [25] Mi Y L, Liu W Q, Shi Y, Li J H. Semi-supervised concept learning by concept-cognitive learning and concept space. IEEE Transactions on Knowledge and Data Engineering, 2022, 34(5): 2429−2442 doi: 10.1109/TKDE.2020.3010918