一种基于互信息的基因数据特征选择算法

A Feature Selection Algorithm for Gene Data Based on Mutual Information

癌症是目前威胁人类生命的重大疾病之一,若能早期发现,则会极大地提高治愈率。针对现在绝大多数特征选择算法都依赖于排除冗余高的特征而没有考虑到特征冗余也可能产生良性影响的问题,笔者提出一种将MIM与CIFE结合的算法。实验结果表明,该算法优于一些其他算法,具有较好的分类精度,对选择出与癌症发病高度相关的基因具有正向作用。

Cancer is one of the most critical major diseases of human life,and if found early,it can greatly improve the cure rate.For the vast majority of feature selection algorithms now rely on excluding the features with high redundancy without considering the possible benign effects of feature redundancy,this paper proposes an algorithm that combines MIM with the CIFE algorithm.The experimental results show that the proposed algorithm outperforms some other algorithms and achieves good classification accuracy and has a positive effect on selecting genes highly r elated with cancer pathogenesis.