Abstract: A method of feature selection for support vector machines is presented. The method is based upon finding those features which minimize. the generalization bound of the SVM and an efficient genetic algorithm is designed to realize it. The method is shown to be effective on both simulation data set and cardiac pattern recognition.

摘要： 提出了一种支撑向量机(SVM)的特征提取方法,该方法使得所提取的特征向量能最小化SVM推广性的界,同时设计了一种有效的遗传算法来实现该方法。模拟数据和心电信号等识别问题的实验结果验证了该方法的有效性。

关键词: 特征选取, 支撑向量机, 推广性的界, 遗传算法

CLC Number: 

• TP301.6