支持向量机集成和特征选择联合算法

Hybrid optimization algorithms of support vector machine ensembles and feature selection

提出了两种基于支持向量机集成和特征选择联合算法。联合算法的核心思想是在构建基础分类器的同时选择有效特征。通过对实测舰船数据和公共数据的识别实验,证明了两种算法都可以用于舰船目标识别。算法一更适用于冗余特征较多的情况。算法二在对舰船目标识别时,选择的特征数目降低为原来特征数目的30%,正确分类率比单个支持向量机高近10%。

Two hybrid optimization algorithms of ensembles of support vector machine （SVM） and feature selection are proposed in this paper. The main idea of these two algorithms is that searching the effective feature subset during the training of base classifier. Experiments are performed on real-world datasets, and the results demonstrate that the proposed algorithms are well suited for ship radiated noise recognition. Algorithm 1 is more suitable for targets whose features are highly redundant. The number of classification features selected by algorithm 2 drops to 30% of the original feature set, meanwhile the classification accuracy increases by about 10% compared to the accuracy of a single SVM classifier.