【作者】 王兴鹏；

【导师】 吕淑然；

【作者基本信息】 首都经济贸易大学 ， 管理科学与工程， 2019， 博士

【摘要】 “情景一应对”模式已经成为当前突发事件应急决策的基本范式。在突发事件应对中,决策者面临时间紧迫性、信息缺失、经验不足等困境,而CBR是能帮助决策者摆脱困境,有效提高决策效率和质量的技术方法。情景是应急决策的基础和依据,因此,情景案例的知识表示则成为“情景-应对”模式下基于案例推理的首要问题,突发事件情景案例涉及的知识丰富、结构复杂,具有跨领域、多粒度、动态关联等特点,而现有的案例表示方法大多把事件案例描述为结构固定、粒度单一、孤立无关的个体,在面对复杂的实际问题时,检索的效率和有效性很低,应用中具有很大的局限性。本文基于粒计算思想,综合运用情景理论、概念格、超图等理论和方法对“情景-应对”模式下基于案例推理的关键技术展开研究。主要研究内容和结论如下:(1)构建了情景案例逻辑结构模型。首先对情景的静态结构进行了分析,明确了突发事件情景的构成要素,总结分析了每类要素的层次结构;然后分析了情景的动态结构,在对情景演化分析基础上,给出情景演化的链式结构;最后综合上述两个方面内容,提出了纵向横向交织的情景案例逻辑结构模型,并提出了适合该结构模型特点的案例表示方法。(2)基于商空间和概念格理论提出了情景案例组织模型。在对商空间和概念格内涵理解基础上,构建了情景商空间和情景概念格模型,实现了对情景案例的划分和组织,完成了对情景案例纵向结构的描述;深入剖析了情景概念格的优点,并提出了情景概念格的构造方法;基于概念提升原理,提出情景概念格粒度变化思想和实现方法。(3)构建了基于超图的情景链表示模型。在对情景演化机理分析基础上,依据超图理论,构建了情景链超图模型,实现了对情景案例横向结构的描述,该模型不但描述了情景演化过程,而且体现了情景事件与关键要素间的相互作用关系;基于情景链超图模型,利用超图拓扑性质提出情景要素风险评价方法;提出了基于情景链超图模型的情景演化规则挖掘方法;基于超边和节点粒化的思想提出情景链超图模型的粒化过程的实现思路。(4)为提高案例检索效率和效果,根据情景案例结构模型的特点,提出了三级检索策略:情景类检索、情景链检索和情景特征检索,该检索过程体现了从粗到细的粒度变化;根据三级检索策略,提出相应的相似度计算方法;提出了应急方案的修正策略和实现思路。(5)基于上述原理和思想设计实现了多粒度情景案例推理原型系统,验证了本文提出模型与方法的有效性性与可行性。该研究为“情景-应对”模式下案例的表示和组织提供了通用框架和模型,为“情景-应对”模式下基于案例推理的实现提供了重要的理论参考,为“情景-应对”应急决策理论研究提供了一种新的思路。更多还原

【Abstract】 The "scenario-response" model has become the basic paradigm of emergency response decisions.In emergency response,decision makers face dilemmas such as time urgency,lack of information or experience.CBR is a technical method that can help decision makers get rid of difficulties and improve decision-making efficiency and quality.Scenario is the foundation and basis of emergency decision-making,so the knowledge representation of scenario case becomes the primary problem of case-based reasoning under the "scenario-response" mode.The cases of emergencies have the characteristics of rich knowledge,complex structure,cross-domain,multiple granularity and dynamic correlation,etc.,and the existing case representation methods mostly describe event cases as individuals with fixed structure,single granularity,and individual isolation.In face of complex practical problems,the efficiency and effectiveness of retrieval is very low,so there are great limitations in this.Based on the above background,this paper combines the theory of granular computing and applies the theory and method of scenario theory,concept lattice and hypergraph comprehensively to study the key techniques based on case-based reasoning in the "scenario-response" mode.The main research contents and conclusions are as follows:(1)Constructed a logical structure model of scenario case.Firstly,analyzed the static structure of the scenario,clarified the components of the emergency scene,and summarized the hierarchical structure of each type of element.Then analyzed the dynamic structure of the situation.By analyzing the scenario evolution process of emergencies,and giving the chain structure of scenario evolution.In the end,combined above two aspects,propose a vertical and horizontal interlaced scenario case logical structure model,and a case representation method suitable for the characteristics of the structure model.(2)Based on quotient space and the concept lattice,a scenario case organization model is proposed.Based on the understanding of the quotient space and concept lattice,the scenario quotient space and the scenario concept lattice are constructed,it can realize the division and organization of the scenario case,complete the description of the vertical structure of the scenario case.It also analyzed the advantages of the situation concept lattice,and proposed the construction method of the situation concept lattice.Based on the principle of concept promotion,the idea and implementation method of the concept of granularity of the concept concept are proposed.(3)A scenario chain model based on hypergraph is constructed.In the foundation of the analysis of the mechanism of scenario evolution,based on the hypergraph theory,the hyperlink model of the scenario chain is constructed,which realizes the description of the horizontal structure of the scenario case.The model not only describes the evolution process of the scenario chain,but also reflects the interactional relationships between scenario event and key elements.Based on the scenario chain hypergraph model,the risk assessment method of scenario elements is proposed by using the topological properties of hypergraphs.The scenario evolution rule mining method based on hyper-profile model of scenario chain is proposed.Based on the idea of super-edge and node granulation,the scenario-chain hypergraph model is proposed.(4)In order to improve the efficiency and effectiveness of case retrieval,according to the characteristics of the scenario case structure model,a three-level retrieval strategy is proposed:scenario category retrieval,scenario chain retrieval and scenario feature retrieval.The retrieval process reflects the granularity change from coarse to fine.According to the three-level retrieval strategy,the corresponding similarity calculation methods are proposed,and also proposed the correction strategy and implementation ideas of the emergency plan.(5)Based on the above principles and ideas,a multi-granularity case-inference prototype system is designed and verified the validity and feasibility of the proposed model and method.The study not only provides a common framework and model for the representation and organization of cases in the "scenario-response" model,but also provides an important theoretical reference for case-based reasoning in the "scenario-response" model.It also provides a new idea for the "scenario-response" emergency decision theory research.更多还原