SCBA:基于演变的规则知识库的数据流分类模型

近年来,数据流作为一种特殊的数据形态,广泛出现于金融、互联网、科学研究等各个领域,引起了数据挖掘界极大的兴趣。其数据量无穷,数据概念随时间变化,对问题要求快速响应等特点给研究工作带来了极大的挑战。该文聚焦于数据流分类问题,提出了一种基于演变的规则知识库的数据流分类模型:SCBA(streamclass ification based on assoc iation ru les)。该模型随着数据流的到达,从动态维护的规则知识库中挑选适应当前数据概念的规则,构建单个分类器。与目前流行的数据流分类模型相比,SCBA具有如下特点:与层次结构的决策树相比,分类器的更新更为灵活并易于实现;与倾向“平均”的组合分类器相比,单个的分类器有利于规避“小模式问题”,并能加快对数据流概念变化的调整速度和对应用的响应速度。

SCBA: An evolving rules-based model for stream data classification

Recently,mining data streams for actionable insights has become an important and challenging task for a wide range of applications in finance,World Wide Web,scientific researches,etc.A rule-based model named SCBA(stream classification based on association rules) was developed for stream data classification based on the evolving rule base.SCBA dynamically selects class association rules from the rule base to update a single classifier to catch the concept drifts in stream data.Compared with the well-known algorithms as decision tree and ensemble classifiers,SCBA has the merits of faster adapting to concept drifts and faster responding to applications.