SIA OpenIR  > 光电信息技术研究室
面向应急救援领域的TCBR方法研究及应用
Alternative TitleTextual Case-Based Reasoning for Emergency Rescue
周习锋1,2
Department光电信息技术研究室
Thesis Advisor史泽林 ; 赵怀慈
ClassificationTP18
Keyword案例复杂度 案例组织 文本案例检索 多专家联合评判 领域知识
Call NumberTP18/Z79/2010
Pages92页
Degree Discipline模式识别与智能系统
Degree Name博士
2010-05-24
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract自1982年Roger C. Schank提出基于案例的推理(Case-Based Reasoning,CBR)思想以来,形成了案例知识表示、经验管理、分布式CBR、Soft CBR等技术,在医学(medicine)、法律(law)、教育(education)、设计(design)、故障诊断(diagnosis)、图像处理(image processing)、推荐系统(recommender Systems)和计划编制(planning)等领域得到了广泛的应用。CBR已成为人工智能领域三大推理方法之一。由于文本案例中存在半结构化或非结构化的问题,建立基于文本案例的推理(Textual CBR,TCBR)系统的难度远大于建立一个CBR系统的难度。本文围绕TCBR系统开发中的案例组织等问题进行了深入的探讨和研究。 1.针对公开的文本案例数据集合中的“问题”和“解决方案”没有明确边界的问题,根据CBR思想的基本原则“相似问题具有相似解”,提出基于复杂度的案例组织方法,据此对半结构化的文本案例进行预处理,确定了文本案例的构成,为以后的各个环节准备了可靠输入。实验结果表明这种方法是可行的、有效的,而且和多位专家的判断结果一致。 2.针对文本案例的问题或解决方案存在多种不同表示的问题,提出了根据复杂度最小原则选择问题空间相似度量的方法,对可选的相似度量进行投票,将得票数最高的相似度量作为问题空间的最佳相似度量。实验结果表明,选出的相似度量具有明显的优势。 3.针对解决方案因为没有类别信息无法定义其正确性的问题,本文提出了基于排序聚合的专家联合决策方法,获得了一个最接近多数专家意见的解决方案序列,将其提交给决策者做最终的选择,由此实现了智能决策支持。这个方法在一定程度上解决了TCBR系统难以评价的问题。 4.针对TCBR系统不能有效利用领域知识的问题,提出了基于Web本体语言(OWL)的领域知识表示,从而在TCBR系统中实现领域知识共享,利用本体表示的领域知识进行描述逻辑推理能够提高TCBR系统问题求解的效率。
Other AbstractSince Roger C. Schank presented the idea of Case-Based Reasoning (CBR) in 1982, there appeared many CBR technologies, such as case knowledge representation, knowledge management, distributed CBR, soft CBR, which have been widely applied in law, medicine, education, diagonosis, image processing, recommender systems and planning, etc. CBR has become one of the three reasoning methods together with rule-based reasoning and model-based reasoning. However, building Textual CBR (TCBR) system is still beyond the capabilities of traditional CBR technologies because textual cases are half-structured data and hard to deal with. This thesis aims to address the issues existed in TCBR systems development. The content of the thesis includes: 1) According to the basic tenet of CBR, similar problems have similar solutions, we put forward the case acquisition method based on textual case complexity in order to prepare textual data set where problem and solution are clarified. This research can offer reliable input for the following processing. Experimental results show that experts agree on the optimizing case pair obtained by the method, which proves that the method is feasible and effective. 2) To reach a consensus among multiple experts feedback about which solution of nearest neghbors is superior to others, rank aggregation method is employed to suggest a unique case ranking to decision-maker. This research achieves intelligent decision support about selecting most similar case. Experimental results show that the challenge on TCBR system evaluation can be conquered. 3) On the basis of multiple experts’ feedback and complexity minimum principle, we propose a voting algorithm among a few candidate similarity measures for the problem space. Experimental results show that it is evident that the selected similarity measure is superior to the other candidates. 4) We design and implement the integration of domain knowledge represented by OWL Web Ontology Language into TCBR systems. Analytical results show that using the technology of description logic reasoning in TCBR can achieve higher efficiency than using the TCBR only.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/9245
Collection光电信息技术研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院研究生院
Recommended Citation
GB/T 7714
周习锋. 面向应急救援领域的TCBR方法研究及应用[D]. 沈阳. 中国科学院沈阳自动化研究所,2010.
Files in This Item:
File Name/Size DocType Version Access License
面向应急救援领域的TCBR方法研究及应用(4131KB) 开放获取LicenseApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[周习锋]'s Articles
Baidu academic
Similar articles in Baidu academic
[周习锋]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[周习锋]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.