Feature selection for multi-class underwater acoustic targets via SVMs and genetic algorithms

In this paper, a new hybrid approach based on support vector machines and genetic algorithms (SVM-GA) is proposed for selecting a feature subset for underwater acoustic targets classification. The genetic algorithm is used for selecting optimal feature subset on the basis of predicted accuracy of SVMs. Four methods are used to predict accuracy of SVMs. This new approach is compared to a classic feature selection called sequential back selection (SBS). The real radiated noises datasets of underwater acoustic targets is considered. The experimental results show that the proposed approach is effective for the purpose of selecting feature subset of underwater acoustic targets.