基于覆盖算法决策界的特征选择算法

Feature Selection Algorithm Based on Decision Boundary of Covering Algorithm

特征选择是模式识别系统的分类器设计之前一个重要而困难的一个课题。在目前现有的方法中,基于决策界的特征选择是其中一类方法。文中将覆盖算法应用于特征提取,提出了基于覆盖算法决策界的特征选择算法(Feature SelectionAlgorithm based on the Decision Boundary of Covering Algorithm,简称FSACA法),然后将该算法应用于一个字符识别的实例并与其他算法比较。实验结果证明了FSACA法的可行性和有效性。

Feature selection is one of the important and difficult subjects before the design of the classifier in a pattern recognition system, Among the existed methods, one kind of these methods is based on the decision boundary. This paper applies covering algorithm into the field of feature selection, and puts forward the Feature Selection Algorithm based on Coveting Algofithm（for short FSACA）, and then a character recognition experiment is done to compare the proposed algorithm with others. The results of the experiment demonstrate the feasibility and the validity of FSACA.