为从不完备决策表中获取最简决策规则,在研究粒计算理论的基础上,提出了一种基于粒计算的决策规则提取算法.该算法利用粒计算的方法对由同一决策表所生成的粒库,从不同的角度和多个层次进行研究,尝试从较低阶的粒库中提取出尽可能多的具有一定规则覆盖度与置信度的简洁规则.实例验证表明,算法能有效地获取最简决策规则,为不完备决策表的最简规则获取提出了一种新方法.
To extract compact decision rules from incomplete decision table, a decision rule extraction al- gorithm based on granular computing was proposed. First, from different point of view and multiple levels, differ- ent granular bases transformed from same decision table were studied. Then, decision rules which satisfy certain rule coverage and confidence limit were extracted as many as possible from relatively low level granular bases. Last, validity for the algorithm was proved by analyzing example.