基于案例推理的应急物资需求预测研究

发布时间：2018-06-06 18:20

本文选题：案例推理 + 相似案例　；参考：《兰州交通大学》2017年硕士论文

【摘要】：近些年,各种自然灾害及突发公共卫生事件频繁发生,为了做好灾后应急救援工作,提高救援效率,最大化降低灾害带来的损失。关键问题是能否在第一时间确定灾区所需的应急物资数量,而后才能将珍贵且有限的物资按时按量的送达受灾人员手中,降低物资浪费,将物资发挥其最大的效用。做好应急物资需求预测工作能给有关部门积极响应和实施救援活动,提供理论依据。本文采用案例推理(Case Based on Reasoning)的方法解决这一问题,运用特定的案例数据作为当前问题解的参照。在决策中用到了大量的形象思维,不是单纯地依靠规则,还考虑到了专家的经验。与传统的基于模型的方法相比,CBR不需要显式表达的领域模型,适于解决不良结构问题。CBR通过调整以往相关问题的成功解来解决新问题。确定案例的表达形式,不同的案例表达形式对整个案例推理过程有不同的影响。本体方法表示突发事件案例时,不仅可以提取数值型的特征属性还可以提取语义型的特征属性,能更详细的描述案例的细节。其中每个特征属性对于整体相似度的贡献用权重来衡量,分别采用层次分析法和核距离法分别确定案例各个属性的权重值,并将不同方法所得的权重值和用不同权重值所确定的相似度进行比较,考察他们的异同。对于添加的语义信息运用基于概念实例的计算方法求解,不仅考虑概念实例之间的距离和概念实例的上层概念还考虑了概念实例所在层次的信息。克服了Wu-palmer算法中对上下层语义拆分计算的不足。整体相似度的值是数值型相似度和语义型相似度的加权和。检索得到的相似案例将为问题案例的求解提供参考依据。在此基础上构建基于本体案例表示和案例推理的应急物资需求预测系统框架。以框架结构为基础,结合应急物资需求领域专业知识,使用Protégé构建应急物资需求领域本体。
[Abstract]:In recent years, a variety of natural disasters and public health emergencies occur frequently, in order to do a good job in post-disaster emergency rescue, improve the efficiency of rescue, maximize and reduce the losses caused by disasters. The key problem is whether the quantity of emergency materials needed in the disaster area can be determined in the first time, and then the precious and limited materials can be delivered to the affected persons on time, so as to reduce the waste of materials and bring the materials into full play. The work of forecasting the demand for emergency materials can provide theoretical basis for the departments concerned to respond positively and carry out rescue activities. In this paper, case based on reasoning (CBR) is used to solve this problem, and the specific case data is used as the reference to the solution of the current problem. A large number of image thinking is used in decision-making, not only relying on rules, but also taking into account the experience of experts. Compared with the traditional model-based approach, CBR does not require explicit representation of the domain model, and is suitable for solving ill-structured problems. CBR can solve new problems by adjusting the successful solutions of previous related problems. Different case expressions have different influences on the whole case-based reasoning process. Ontology method can not only extract the numerical feature attribute but also the semantic feature attribute, and describe the details of the case in more detail. The contribution of each feature attribute to the overall similarity is measured by weight, and the weight of each attribute is determined by the analytic hierarchy process (AHP) and the kernel distance method, respectively. The weights obtained by different methods are compared with the similarity determined by different weights, and their similarities and differences are investigated. The added semantic information is solved by the method based on the concept instance, which not only considers the distance between the concept instances and the upper concept of the concept instance, but also considers the information of the level where the concept instance is located. It overcomes the shortcomings of Wu-palmer algorithm for semantic splitting of upper and lower layers. The value of global similarity is the weighted sum of numerical similarity and semantic similarity. The similar cases retrieved will provide a reference for the solution of problem cases. On this basis, the framework of emergency material demand prediction system based on ontology case representation and case-based reasoning is constructed. Based on the framework structure and combining the expertise in the field of emergency material demand, Prot 茅 g 茅 is used to construct the ontology of emergency material demand domain.
【学位授予单位】：兰州交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【参考文献】