朴素贝叶斯分类器一阶扩展的注记

Notes on First-order Upgrading to Nave Bayesian Classifier

众多研究者致力于将朴素贝叶斯方法与原有的ILP系统结合,形成各种各样的多关系朴素贝叶斯分类器(MRNBC)。该文提出形成朴素贝叶斯分类器的一阶扩展的一般方法。现实中关系数据库广泛存在,可以直接作用于数据库表,而无须转换表示形式的MRNBC则是研究的重点,该方法主要基于关系数据库理论,分析了进行一阶扩展的关键问题。

Many researchers focus on the combination of ILP system together with the naive Bayesian classifier. And various Multi-Relational Naive Bayesian Classifiers（MRNBC） have been proposed. This paper describes a methodology based on ILP for upgrading naive Bayesian classifiers to first-order logic. There are almost relational databases everywhere in real-life, so this paper also aims at solving how to shape MRNBC based on relational databases theory. The upgrading can deal with data in multiple tables directly and transformation is not needed.