基于模糊MEBN的不确定性本体表示和推理研究

发布时间：2018-06-13 20:30

本文选题：本体 + OWL　；参考：《郑州大学》2017年硕士论文

【摘要】：随着语义网中数据的不断丰富和语义服务的不断发展,语义网中开始出现大量的不确定数据,给语义网的应用带来很大挑战,不确定性数据的表示和推理变得更加重要。本体能够对语义网中的语义和知识进行建模,但现有的本体语言无法直接表示不确定性知识,需要对本体进行扩展。目前扩展方式主要基于概率和基于模糊两方面,但现有的研究往往只关注于其中一方面,而在实际应用中两者可能同时出现。针对现状,本文对不确定性的表示和推理的研究进行分析和总结,并对模糊概率知识的表示和推理进行研究,提出基于模糊多实体贝叶斯网络(模糊MEBN)的本体表示和推理的框架。本文主要工作如下:首先,对语义网中不确定性知识的表示方法进行研究,将模糊MEBN理论和本体相结合,提出基于OWL2(Web Ontology Language)的模糊MEBN本体语言FuzzyPR-OWL。该本体语言通过OWL2语言构建能表示模糊概率知识的本体类和属性,提供对模糊语义和不确定关系的描述,并给出语法定义和语义解释,同时用实例说明FuzzyPR-OWL本体构建领域本体的方法。之后,对不确定性知识的推理方法进行研究。基于FuzzyPR-OWL本体表示提出模糊概率本体的推理框架。论文结合模糊概率和贝叶斯网络的信念传播算法,在节点间传播的消息中增加模糊规则的影响因素,提出基于模糊概率的信念传播算法,在此基础上给出推理过程,包括数据的模糊化、SSFBN的构建以及模糊信念传播。最后,通过实验验证模糊概率本体的表示和推理方法的可行性和有效性。先用所提出的方法对汽车防撞警告系统进行建模和推理,把增加模糊状态后的推导结果与无模糊的概率推导结果作对比,得出模糊状态对目标节点概率的影响结果,然后在数据集上利用十倍交叉验证法对算法准确性进行评估。实验结果表明,本文提出的本体语言FuzzyPR-OWL能够有效表示和推理模糊概率知识,为不确定性信息的表示和推理提供一种新的解决方案。
[Abstract]:With the continuous enrichment of data and the development of semantic services in the semantic Web, a large number of uncertain data appear in the semantic Web, which brings great challenges to the application of the semantic Web, and the representation and reasoning of uncertain data becomes more important. Ontology can model semantics and knowledge in semantic web, but the existing ontology language can not express uncertain knowledge directly, so ontology needs to be extended. At present, the methods of expansion are mainly based on probability and fuzzy. However, the existing researches focus on only one of them, which may occur simultaneously in practical applications. In view of the present situation, this paper analyzes and summarizes the representation and reasoning of uncertainty, and studies the representation and reasoning of fuzzy probability knowledge. A framework for ontology representation and reasoning based on fuzzy multi-entity Bayesian network (fuzzy MEBN) is proposed. The main work of this paper is as follows: firstly, the representation method of uncertain knowledge in semantic web is studied. The fuzzy MEBN ontology language FuzzyPR-OWL based on OWL2 Web Ontology language is proposed by combining fuzzy MEBN theory with ontology. The ontology language constructs ontology classes and attributes that can represent fuzzy probability knowledge through owl 2 language, provides descriptions of fuzzy semantics and uncertain relations, and gives syntax definition and semantic interpretation. An example is given to illustrate the method of constructing domain ontology by FuzzyPR-OWL ontology. Then, the reasoning method of uncertain knowledge is studied. A fuzzy probability ontology reasoning framework is proposed based on fuzzy PR-OWL ontology representation. Based on fuzzy probability and belief propagation algorithm of Bayesian network, this paper proposes a belief propagation algorithm based on fuzzy probability by adding the influence factors of fuzzy rules to messages propagated between nodes, and then gives the reasoning process. It includes the fuzzification of data and the construction of SSFBN and the propagation of fuzzy belief. Finally, the feasibility and validity of the representation and reasoning method of fuzzy probability ontology are verified by experiments. First, the vehicle anti-collision warning system is modeled and inferred by the proposed method, and the results after adding fuzzy state are compared with the result of non-fuzzy probability derivation, and the effect of fuzzy state on the probability of target node is obtained. Then the accuracy of the algorithm is evaluated by using ten times cross validation method on the data set. Experimental results show that the proposed ontology language FuzzyPR-OWL can effectively represent and infer fuzzy probability knowledge and provide a new solution for the representation and reasoning of uncertain information.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.1;TP18

【参考文献】