多属性群决策的多粒度三支建模理论与方法

Multi-attribute group decision making is an efficient approach for multi-variable complicated decision making problems by integrating collective wisdom. The intensified trend of social datalization has promoted the rapid development of data modeling, and permeated in the field of decision making. Aiming at fusing granular computing with three-way decisions to study scheme synthesis and analysis of solution space in multi-granularity levels, multi-granularity three-way modeling provides multi-dimension problem solving methods for multi-attribute group decision making problems. The project is intended for the need of multi-attribute group decision making and comprehensively considers the uncertain, multi-level, multi-strategy and multi-period decision making characteristics, in allusion to basic key scientific problems of multi-granularity three-way modeling such as granulation uncertainties, decision making risk preferences, optimal granularity selections, dynamic decision making characteristics, etc., the project studies and develops multi-granularity three-way modeling theory and method of multi-attribute group decision making. The main contents include: ⑴ Generalized fuzzy multigranulation decision-theoretic rough set models over two universes with decision making risk preferences; ⑵ Generalized fuzzy multigranulation decision-theoretic rough set models over two universes based on optimal granularity selections; ⑶ Generalized fuzzy multigranulation rough set models over multiple universes and generalized fuzzy multigranulation decision-theoretic rough set models over multiple universes with dynamic decision making characteristics. The anticipative research results will not only innovate multi-attribute group decision making theories and methods, but also deepen multi-granularity three-way modeling theories and methods, which are equipped with important scientific significances and potential application values.

多属性群决策是一种集群体智慧应对多变量复杂决策问题的有效方法。社会数据化的加剧趋势促进了数据建模快速发展，并渗透于决策领域。多粒度三支建模旨在通过融合粒计算与三支决策研究多粒度层次上解空间的方案集成与分析，可为多属性群决策问题提供多维度问题求解方法。本项目面向多属性群决策需求，综合考虑不确定、多层次、多策略、多阶段等决策特点，针对多粒度三支建模的粒化不确定性、决策风险偏好、最优粒度选择、动态决策特性等基础性核心科学问题，研究发展多属性群决策的多粒度三支建模理论与方法。主要内容包括：⑴考虑决策风险偏好的双论域广义模糊多粒度决策粗糙集模型；⑵基于最优粒度选择的双论域广义模糊多粒度决策粗糙集模型；⑶考虑动态决策特性的多论域广义模糊多粒度粗糙集模型和多论域广义模糊多粒度决策粗糙集模型。项目成果既可创新多属性群决策理论与方法，又可深化多粒度三支建模理论与方法，具有重要科学意义与潜在应用价值。

随着云计算、物联网、工业互联网等产业的快速发展，社会经济发展中各类复杂决策问题不断涌现。多属性群决策是一种集群体智慧应对多变量复杂决策问题的有效方法。社会数据化的加剧趋势促进了数据建模快速发展，并渗透于决策领域。. 本项目针对经典决策数据建模在求解具备多粒度特征的多属性群决策问题时遇到的挑战，聚焦于粒化不确定性、决策风险偏好、最优粒度选择、动态决策特性等多粒度三支建模基础性核心科学问题，结合粒计算框架下的多粒度三支建模方法，深入研究多属性群决策的多粒度三支建模理论与方法。.项目取得主要研究成果如下：（1）在多粒度三支建模中的决策风险偏好刻画方面，提出了基于决策风险排序与调节机制的多粒度决策粗糙集模型、考虑信息融合方法稳定性以及决策风险偏好的多粒度概率粗糙集模型，并在人岗匹配、课堂教学质量评价等背景下提出了多属性群决策算法。（2）在多粒度三支建模中的最优粒度选择与动态决策刻画方面，提出了研究了基于逼近理想解法与集成算子法的多粒度决策粗糙集模型、区间犹豫模糊图模型，并在并购对象选择、空气质量评价等背景下提出了多属性群决策算法。（3）在多粒度三支建模中的决策蕴涵推理方面，提出了决策蕴涵规范基生成算法，基于逻辑公式定义了逻辑型决策蕴涵，设计了逻辑型决策蕴涵的语义框架，给出了模糊决策蕴含在模糊形式背景下的解释。（4）在多粒度三支建模中的多标记特征选择方面，提出了基于相对辨识对矩阵的多标记特征选择算法、提出了基于双空间模糊辨识关系的多标记特征选择算法。. 项目成果既可创新多属性群决策理论与方法，又可深化多粒度三支建模理论与方法，具有重要的科学意义。项目成果将应用于山西省科技信息匹配与知识发现与协同办公集约化平台领域，并进行成果转化。