中文摘要：

从信息论的角度分析目前属性约简算法的研究成果，并指出它们关于冗余处理显得不够完善；然后利用条件属性间相对独立性特点，划分出等价类，并以此为基础提出一种基于粗糙等价类的属性约简算法．最后对顾客评价表中部分数据进行了实例分析，以此表明该算法能够在保留原始有用数据的基础上有效减少冗余，减小误差，得到良好的效果．

英文摘要：

This paper analysis the latest research results regarding attributes reduction from the perspective of information theory, and pointed out that method is not prefect on the redundancy. After that the paper use the characteristics of condition attributes which rela- tive independence with each other to divide into equivalence class, and on that basis propose an attribute reduction algorithm of equiva- lence class that based on rough sets. At last,we do the sample analysis the partial datas of customer~ evaluation table in order to indicate this algorithm can reduce the redundancy and reduce the errors effectively and can achieve a good performance without destroy the usefulness of the primary datas.