基于DL关联ε■++规则挖掘的归纳知识发现

DL associationε■++rules mining based inductive knowledge discovery

为了从不完整和动态的数据中发现知识,提出了一种基于DL关联ε■++规则和归纳推理的一致知识发现。首先通过对描述逻辑ε■++规则和演化本体的知识动态性地分析得到了演化本体中的归纳推理学习,它是基于原子集支持度和权值以及关联ε■++规则的置信度,通过挖掘ε■++规则来实现的;其次,通过获得具有最小支持度和最小权值的代表性关联DLε■++规则,实现对重要规则的精确识别,从而实现归纳知识发现。采用来自于某市历史数据的实验结果表明,提出的方法相比于现有的主流方法在演化本体和动态语义数据中的知识发现不仅有很好的扩展性,而且有更高的准确性。

In order to discover knowledge from incomplete and dynamic data,this paper proposed a technique for consistent knowledge discovery based on DL associationε■++rules and inductive reasoning.Firstly,by analyzing description logicε■++rules and the dynamics of the knowledge of evolving ontology,it obtained inductive reasoning learning in evolving ontology.Basing on the support and weight of atomic set and the confidence of associationε■++rules,it achievedε■++rules mining.Secondly,by getting the representative association DLε■++rules with minimum support and minimum weight,it realized the precise identification of fundamental rules so as to complete inductive knowledge discovery.The experimental results for historical data from a certain city show that,compared with the existing mainstream methods,the proposed technique not only has better scalability,but also has higher accuracy in terms of evolving ontology and dynamic semantic data knowledge discovery.