多粒度语言词计算理论及其在“情景-应对”型应急决策中的应用

Computing with Words(CWW) is very important for overcoming traditional computing paradigm’s limitations. CWW has been used to some kinds decision problems for reducing the gap between human reasoning and numerical computing. Based on the methods of fuzzy mathematics, operations research, artificial intelligence and computer simulation, this project proposes a multi-granular interval linguistic computing model based on 2-tuple and uncertain linguistic models, according to multi-granular linguistic computing problem. We use interval linguistic variables as the uniform for different vagueness．And we propose the transformation method between multi-granular linguistic variables. Establish operation rules and methods of fuzzy linguistic function, including basic arithmetic operation, vector comparison, distance operation, etc. According to the shortcoming of 'Scenario-response' emergency decision theory to deal with fuzzy sophisticated scene, this project uses multi-granular linguistic variables to build scenario prototype. This scenario prototype can describe decision makers' feeling with social and natural environments. The CWW theory is used to deeply analyze the 'Scenario-response' emergency decision paradigm and rule, and is helpful to build fuzzy reasoning rule based on fuzzy Petri net. So we can find out solutions for emergency events quickly. This application can help to understand and grasp the process of emergency events from happening to development and evolution. So it is meaningful for improving emergency management level.

词计算理论对克服传统计算范式的局限性具有重要意义，已经被广泛应用于各类决策问题，用以缩小人类推理与计算之间的差距。本课题将基于模糊数学、运筹学、人工智能和计算机仿真等方法，针对多粒度语言计算问题，结合二元语义、不确定语言模型，提出多粒度区间语言的符号化计算模型。用区间语言为不同模糊程度的偏好建立统一形式，实现多粒度语言之间的转换。提出多粒度语言变量的运算规则和计算方法，包括基本算术运算、向量比较、相似度和距离运算等。针对“情景－应对”型应急决策范式在模糊复杂情景问题处理中的不完备之处，本课题采用多粒度语言构建情景原型，用来描述决策者对情景的社会环境和自然环境的感知。并采用词计算理论深入分析“情景－应对”决策范式与决策规则，形成基于模糊Petri网的逻辑推理，为突发事件快速找到应对方案。本应用有助于理解和把握突发事件从发生到发展再到演化的过程，对提高应急管理水平均具有一定意义。

词计算理论对克服传统计算范式的局限性有重要意义，已经被广泛应用于各类决策问题，来缩小人类推理与计算之间的差距。本课题主要研究内容有三个方面：一是结合多粒度语言计算模型，提出多种基于不同偏好形式的模糊决策方法；二是提出模糊复杂情景下，决策者以社会感知为偏好的“情景－应对”决策范式与决策规则；三是提出非常规突发事件快速生成解决方案的方法。.重要结果和科学意义如下：.1、提出一系列多粒度语言的距离测度计算方法，用于集结不同偏好和衡量群决策的一致性。新的距离测度在多粒度语言的一致性优化和避免信息丢失上表现出优越性。.2、结合图形理论和符号化方法提出多种非均衡多粒度语言计算模型，解决群决策和TOPSIS问题。证明了新的方法比欧氏距离测度等其他测度更有优越性。.3、通过对非常规突发事件的应急知识管理分析，给出了多种基于案例推理的应急决策方法。这些方法解决了“情景-应对”型决策的应急知识利用效率问题，对非常规突发事件快速生成解决方案具有较大的参考价值和现实意义。.4、应用二型模糊数的计算方法，建立基于模糊群决策的需求预测模型。该模型能在信息匮乏条件下预测应急物资需求，能保留更多的个体预测信息。