CAJ | 학술논문

模块化分类规则的归纳学习是机器学习领域应用较广的一类算法，已经发展形成了多个分支和派系，其中，Prism 系列算法是当前该类学习算法的典型代表之一。Prism 系列算法历经近20多年的发展，在多种归纳学习任务中得到了成功应用，目前已经成为决策树归纳算法的主要竞争者。本文在 Prism 系列算法基本框架的基础上，分别从单机算法和并发算法两个方面对 Prism 系列算法进行了综述，比较分析了不同算法对于多种分类问题的适应性、优缺点及相互关系，并展望了未来该类算法的发展方向。
모괴화분류규칙적귀납학습시궤기학습영역응용교엄적일류산법，이경발전형성료다개분지화파계，기중，Prism 계렬산법시당전해류학습산법적전형대표지일。Prism 계렬산법력경근20다년적발전，재다충귀납학습임무중득도료성공응용，목전이경성위결책수귀납산법적주요경쟁자。본문재 Prism 계렬산법기본광가적기출상，분별종단궤산법화병발산법량개방면대 Prism 계렬산법진행료종술，비교분석료불동산법대우다충분류문제적괄응성、우결점급상호관계，병전망료미래해류산법적발전방향。
Inductive learning with modular classification rules is a kind of widely used algorithms in machine learning.It has developed in various aspects and branches,and one of its modern representatives is the Prism family of algorithms.The Prism family of algorithms has been applied in many kinds of inductive learning tasks after more than twenty years development,and it has become a major competitor to the induction of decision trees.This paper gives a survey on the Prim family of algorithms that run independently and simultaneously, respectively,based on the basic framework of Prism algorithms.The research has made a comparison study on multiple algorithms,with respect to their suitability,advantages and interrelation in multiple classification tasks.Furthermore,the perspectives of modular classification inductive learning algorithms are presented.