基于序优化方法的特征选择研究

On the applications of ordinal optimization in feature selection

针对模式识别中特征子集的选择存在组合爆炸的情况,以核爆地震识别中的特征选择为例,将序优化中的BP(Blind Picking)规则进行用于特征选择研究。实验结果表明,此优化方法能迅速地减小搜索空间,提高搜索效率,从而为特征选择问题提供了一种新的处理方法。

In this paper, a new method of feature selection based on ordinal optimization is presented, and an experiment based on BP rule is carried out to select the optimum feature set of underground nuclear explosion and natural earthquake. Experimental results indicate that this method can reduce the search space and enhance the search efficiency effectively. This may be a novel theoretical method in feature selection.