Abstract: One of rough set theory essence is knowledge reduction. A kind of knowledge relative reduction algorithm based on GA was proposed. Conditional information entropy was used to compute relevance of attributes and it was used in fitness function to assure reduction has fewer attributes and relevance of attributes. The test results by UCI Machine Learning Data Set show this algorithm is effective.

摘要： 针对粗糙集理论核心内容之一的知识约简问题,提出了一种基于遗传算法的粗糙集属性约简算法。利用条件熵计算属性间的相关性,并将其引入到适值函数中,可以保证所求约简含有较少的属性而且属性间的相关性较小。实验证明,它可以得到比较理想的结果,对UCI机器学习数据集的测试结果也验证了算法的有效性。