本体概念分类的遗传算法  

An ontology concept classification method based on genetic algorithm

作　　者：王俊丽[1] 丁志军[1,2] 蒋昌俊[1] 孙萍[1] 

机构地区：[1]同济大学嵌入式系统与服务计算教育部重点实验室,上海201804 [2]山东科技大学信息科学与工程学院,青岛266510

出　　处：《高技术通讯》2008年第3期265-270,共6页Chinese High Technology Letters

基　　金：上海市基础研究重点项目(05JC14063);上海市科委2006年度“登山行动计划”(06JC14065);上海市国际科技合作基金项目(062107037)资助

摘　　要：提出了一种基于遗传算法的本体概念分类规则学习方法(GAOCL)。该方法从已有的本体库或知识库中获取实例作为训练样本,通过遗传算法获取相关的概念分类规则,实现概念实例的有效分类,以达到扩充和丰富本体的目的。首先,在编码过程中采用了可变长度的规则集合作为个体,以满足不同概念对分类规则数目的不同要求。其次,定义了规则集相对覆盖率,并以此作为适应值函数,评估个体对概念实例的分类效果,实现优化迭代。最后,给出了基于遗传算法的本体分类规则学习算法。利用一组标准样本集对该算法与同类算法进行了比较,实验结果表明该算法具有很好的收敛性,并且能获得较好的分类效果。This paper presents the GAOCL, a genetic algorithm （GA） based method for learning ontology concept classification rules, by which a set of rules relative to concept classification can be achieved by GA to classify the concept examples from an ontology for the purpose of expanding and enriching the ontology. In the learning process, on the one hand, each individual represents an unordered set of fixed-length rules and the length of individual in the population is variable. On the other hand, the fitness function is defined as relative coverage rate of a set of rules, which can guarantee that the achieved rules can cover as many as possible instances of a given concept, and cover as few as possible instances of other concepts. The experimental results show that the algorithm has good convergence and can obtain optimal rule set.

关 键 词：本体学习 遗传算法 概念分类规则 收敛性 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]