生物信息学中的特征选择

摘要

We summarize feature selection approaches with recent applications in bioinformatics.We suggest to deviate from the commonly used categorization of feature selection approaches into filter,wrapper,and embedded approaches.Instead,we view feature selection as a combinatorial optimization or a search problem,by classifying feature selection approaches into exhaustive search,heuristic search,and hybrid methods,with heuristic search approaches further divided into those with or without feature importance ranking.%回顾了特征选择的主要原理及其在生物信息学中的最新应用.我们将特征选择看作组合优化或搜索问题,将特征选择法分为穷举搜索法、启发式搜索法以及混合法,其中启发式搜索法可以被进一步分为是否结合数据特征重要程度的排序,这样比常规对特征选择方法以滤波、封装和嵌入式的分类更为合理.