摘要
作为一种近似处理的工具,粗集主要用于不确定情况下的决策分析,并且不需要任何事先的数据假定。但当前的主流粗集分类方法仍然需要先经过离散化的步骤,这就损失了数值型变量提供的高质量信息。本文对隶属函数重新加以概率定义,并提出了一种基于Bayes概率边界域的粗集分类技术,比较好地解决了当前粗集方法所面临的数值型属性分类的不适应、分类规则不完备等一系列问题。
Having been broadly used in decision-making fields Rough Set Theory（RST）provides a way of extracting decision rules without imposing apriori assumptions.However current RST-based classification methods still need to discrete numerical variables into categorical ones,in which potential useful information may be omitted.In this article,we introduce a Bayes-based RST classification technique which can solve a series of problems facing with current RST classification,including inability to numerical data,incomplete rule generation and etc.
出处
《统计研究》
CSSCI
北大核心
2010年第3期76-82,共7页
Statistical Research
