个体化语言信息特征表示与聚合范式及其群体决策模型

Linguistic information in group decision-making processes frequently exhibits an inherent level of individual heterogeneity. The present project aims at developing a novel suite of character representation models and aggregation paradigms for personalized linguistic information. The benefits of individually modeling the character representation for given linguistic information and putting forward aggregation paradigms for meaningfully combining it will be integrated into the context of group decision-making. For this, the present project is to introduce a number of personalized linguistic group decision-making models that will be defined for various decision frameworks, based on the previously devised representational models and aggregation strategies. The detailed specific objectives of the research undertaken within the scope of this project are as follows. The existing rule of approximately equivalent transformation for comparative linguistic expressions will be reformulated on the basis of subjective probabilistic linguistic term sets. The personalized linguistic information initially provided by the individuals can then be modeled from the perspective of individual participation. Likewise, an adjustment algorithm driven by the attitude change of individuals will be designed for obtaining attitudinal personalized linguistic information associated with such attitudinal changes describing individuals’ behaviors. The challenge of “combinational explosion” raised from the aggregation of subjective probabilistic linguistic term set is pointed out, and in order to tackle it, the aggregation theory is adapted to the development of subjective probability distribution combination approaches, and based on which a novel aggregation paradigm is proposed for personalized linguistic information. Following the strategy of decomposing the aggregation process into two stages, an additional aggregation paradigm for personalized linguistic information is introduced on the basis of clustering analysis techniques. Subsequently, different weighting approaches for assigning different degrees of importance to the participants and decision criteria with various interrelationships among them are developed. Importantly, the aggregation and subsequent decision-making processes are conducted in full accordance with the proposed character representation models and aggregation paradigms for personalized linguistic information. As such, several personalized linguistic group decision-making models will be further constructed to accommodate various decision frameworks and practical real-world situations. A series of simulation-based experiments and cases of study will be conducted to verify and validate the practical applicability of the proposed decision-making approaches. The research outputs of this project will serve as the theoretical ground for the manipulation of personalized linguistic information mining and computing in group decision-making contexts.

以群体决策场景下语言信息的个体性差异特征为对象，研究语言信息的个体性特征表示与聚合范式，提出个体化语言信息特征表示模型及其新型聚合范式。研究的主要内容为：以主观概率语言术语集为载体，制定比较性语言短语的近似等价转换规则，提出考虑个体参与度的初始个体化语言信息特征表示模型，设计决策个体态度转变驱动的语言信息个体化调整算法；针对既有主观概率语言术语集聚合过程中存在的“组合爆炸”问题，提出基于聚合函数理论的主观概率分布合成方法，构建个体化语言信息基于主观概率分布合成的聚合范式；制定两阶段聚合过程分解策略，提出基于聚类分析方法的个体化语言信息聚合范式；以个体化语言信息特征表示模型和聚合范式为基础，提出面向具有异化关联关系的决策个体与准则的赋权方法，建立个体化语言信息群体决策理论模型和设计仿真实验模型，给出个体化语言信息群体决策规则和仿真结果分析。成果将为群体决策中语言个体性挖掘与计算提供理论依据。

本项目旨在研究群体决策场景下语言信息的个体性特征表示与聚合范式及其群体决策模型，现将所取得的研究成果总结为以下三个部分：（1）个体化语言信息特征表示模型与算法设计：构造了基于语言相似度的HFLTS概率分布计算结构，建立了个体化语言的态度特征模型；以主观概率语言术语集为载体,在比较性语言短语向HFLTS转换的过程中引入态度特征参数，提出了评估态度驱动的个体化语义转换方法，以获得嵌入评估态度的准确语言结果；研究了初始个体化语言信息特征表示模型的态度特征参数确定方法和程度估计方法，提出了态度转变下主观概率分布信息完全和不完全情境的调整模型优化目标和约束，设计了基于态度转变的个体化语言信息特征表示调整算法，并就各类态度转变调整算法在生成主观概率分布估计的一致性程度进行比较性试验。（2）个体化语言信息的聚合范式开发：将基于HFLTS的主观概率分布概念重构为基于H2TLTS的概率分布，结合可能性二元语言对(P2TLP)的概念架构，提出了基于HFLTS的主观概率分布的两阶段聚合范式；提出了基于相似度测度的聚类分层聚类(SM-AggHC)算法，减少了所考虑的聚合集的基数，为现有基于主观概率语言术语集聚合过程中出现的“组合爆炸”问题提供解决思路；在此基础上，运用不确定语言信息基本单元的信息测度工具，分析了各聚类算法在对群体语言信息降势过程中的时间与空间复杂度，提出了在运行时间和运行内存上均优于SM-AggHC算法的K-means聚类驱动的n -2阶段聚合算法(N2S-KMC)。（3）面向建筑材料优选问题的基于个体化语言信息的群体决策模型构建，提出了基于个体化语言信息的建材优选群体决策模型，为建材优选问题提供了新的方法论。项目资助下，申请人入选湖北省楚天学者计划，受邀担任INS、SOCO等领域内权威期刊副主编。共发表论文51篇，其中被SCI/SSCI检索40篇次，EI检索9篇次，1篇论文发表在工程管理领域顶尖期刊AUTCON上，入选ESI热点论文1篇次，高被引论文4篇次。培养毕业硕士研究生4名，在读硕士研究生6名。培养学生获得研究生国家奖学金、武汉大学优秀毕业生、武汉大学学术创新奖二等奖、武汉大学光华奖学金等荣誉14次。