基于知识元和模糊认知图的应急案例推理研究

发布时间：2018-07-12 11:03

本文选题：模糊认知图 + 知识元　；参考：《大连理工大学》2015年硕士论文

【摘要】：案例推理可以根据以往经验为突发事件的应急管理提供决策辅助,当前案例推理的相关研究中大致可以归为两类,其中最主流的方法是基于案例推理的方法(CBR),大量学者从案例属性、特征的角度出发提出了案例结构表示、案例检索匹配方法等,对CBR方法做了众多的改进型研究。另外,有部分学者尝试着从案例的内容出发,针对特定领域内的应急突发事件的结果进行模拟或预测。这两类方法普遍存在的问题有：一是现有CBR类方法从案例属性、案例特征的角度出发通过类比推理的思想来解决问题,须要人为定义案例结构及案例表示方法,从而导致领域概念、案例结构无法统一表示；二是现有基于案例内容的推理方法处于探索阶段且相关研究相对较少,其推理过程相对简单,推理结果需要专家参与解释；另外,现有两类方法缺乏对事件要素间的隐性知识和应急突发事件的发展过程的研究,且人为影响因素占比很大导致推理结果缺乏客观性。针对上述局限,本文以应急案例作为研究对象,通过挖掘事件要素间的关联信息和因果知识来构建模糊认知图(FCM)从而进行案例推理。首先,通过文本挖掘的方法获取应急案例中的领域概念知识元,利用关联分析方法抽取领域知识元间内在的联系,挖掘领域概念间的隐性知识,即事件要素间的因果关联知识；然后,根据所获取到的领域知识元间的因果关联知识来识别FCM结构的节点及节点间的影响权重,从而构建了应急案例的模糊认知图推理模型,通过连接事件要素来统一表示案例结构及领域概念,并分析了应急案例FCM的推理过程及结果；最后,选择煤矿瓦斯类应急案例文本作为实验对象,抽取瓦斯类应急案例的事件要素并挖掘出要素间的因果知识,利用本文模型构建了煤矿瓦斯类应急突发事件的FCM推理模型,推理结果证实了本文方法的实用性及可靠性。本研究提出了基于领域知识元的应急案例FCM推理模型有其优势：一是,从基于案例内容的角度出发,利用领域概念知识元作为应急案例FCM的节点可以表示出事件要素间内在的因果知识,避免了人为影响因素过多的弊端,解决了现有方法从案例属性、案例特征出发无法统一表示案例结构、领域概念的问题；二是,应急案例FCM推理模型的推理过程可以展示出整个事件的演化过程及各个事件要素的状态变化过程,相比现有案例推理方法提高了案例推理的准确度和客观性。另外,利用应急案例FCM推理模型可以为决策者掌握事件发展状态、预测事件结果、引导事件发展走向等提供决策依据。
[Abstract]:Case-based reasoning (CBR) can provide decision support for emergency management according to past experience. The most popular method is case-based reasoning (CBR). A large number of scholars put forward case structure representation and case retrieval matching method from the point of view of case attributes and features. In addition, some scholars try to simulate or predict the results of emergency emergencies in a specific field. The common problems of these two methods are as follows: first, the existing CBR methods solve the problems by analogy reasoning from the view of case attributes and case characteristics, and need to define the case structure and case representation methods artificially. As a result, the domain concept and case structure can not be uniformly represented; second, the existing case-based reasoning methods are in the exploratory stage and the relative research is relatively few, the reasoning process is relatively simple, and the reasoning results need experts to participate in the interpretation; in addition, There is a lack of research on the tacit knowledge between the elements of events and the development process of emergency emergencies, and the large proportion of human factors leads to the lack of objectivity of reasoning results. In view of the above limitations, this paper takes emergency cases as the research object, and constructs fuzzy cognitive map (FCM) by mining the correlation information and causality knowledge between event elements to carry out case-based reasoning. Firstly, the domain concept knowledge element in emergency case is acquired by text mining method, and the intrinsic relation between domain knowledge element is extracted by association analysis method, and the tacit knowledge among domain concepts is mined, that is, causal association knowledge between event elements. Then, according to the knowledge of causality relation between domain knowledge elements, the nodes of FCM structure and their influence weights are identified, and the fuzzy cognitive graph reasoning model of emergency case is constructed. The case structure and domain concept are unified by connecting event elements, and the reasoning process and results of emergency case FCM are analyzed. Finally, the text of coal mine gas emergency case is selected as the experimental object. This paper extracts the event elements of gas emergency cases and excavates the causality knowledge between them. The FCM reasoning model of coal mine gas emergency emergencies is constructed by using the model of this paper. The result of reasoning proves the practicability and reliability of the method in this paper. In this study, the FCM reasoning model based on domain knowledge element has its advantages: first, from the perspective of case-based content, Using domain concept knowledge element as the node of emergency case FCM can express the inherent causality knowledge among the elements of the event, avoid the malpractice of too many human factors, and solve the existing method from the case attribute. The problem of case structure and domain concept can not be represented by case characteristics. Second, the reasoning process of emergency case FCM reasoning model can show the evolution process of the whole event and the state change process of each event element. Compared with the existing Case-Based reasoning (CBR), the accuracy and objectivity of Case-Based reasoning (CBR) are improved. In addition, the emergency case FCM reasoning model can provide decision basis for decision-makers to grasp the state of event development, predict the outcome of events, and guide the development of events.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：D035.2

【参考文献】