一种变粒度的规则提取算法

属性约简和值约简是粗糙集理论中知识获取的重要组成部分。通常，在知识获取的过程中先进行属性约简，然后在其基础上进行规则提取。但在实际应用中，属性约简在简化信息系统与提高规则提取效率的同时，原始信息系统中有些重要的条件属性可能被丢弃，从而导致属性约简后对信息系统进行知识获取得到的规则其数量与简化程度并不占优。针对上述问题，提出一种基于粒度变化的规则获取算法，通过属性粒度从粗到细的变化，直接从原始信息系统中提取规则；采用该方法得到的规则与属性约简后得到的规则相比，它们的数量与平均每条规则包含的特征属性数相对较少。最后，在理论分析的基础上，通过实例验证了算法可行性，并通过实验验证了算法的正确性和高效性。

An algorithm for rule extraction based on changing granularity

The attribute reduction and value reduction are two important parts of knowledge acquisition in rough set. Generally, the method of knowledge acquisition is based on rule extraction of attribute reduction. But in the actual application,attribute reduction simplifies the information system and improves rule extraction''s efficiency. At the same time, some important condition attributes may be discarded. So it has the direct result that the numbers and simplification of rules are not necessarily good. Therefore, in this paper, an algorithm based on changing granularity is presented, and with this method we can obtain rules step by step when the granularity is changed from coarse to fine. The rules can extracted from the original information system. Compared with the results after attribute reduction, the numbers and simplification of rules are better. Theoretical analysis and experimental result show that the algorithm of this paper is feasible. Finally, experimental results show that the new algorithm is not only accurate but also efficient.