A GA-based SVM Decision-tree Multi-Classification method

Recently, in the fields of machine learning, how to use support vector machine for multi-class objects classification while improving the classification efficiency of the classifier has become one of the main study points, effective solutions to this problem have great significance for improving the probability of target recognition. In this paper we present a GA-based SVM decision tree algorithm. In our algorithm, we randomly generate a decision tree to build the SVM classifier on the same test samples of the classification accuracy rate as the genetic algorithm fitness function, then with the help of genetic algorithm,we can find the optimal decision tree, and then construct an optimal decision tree SVM classifier as the optimal SVM classifier. We use this algorithm to deal with the low altitude flying passive acoustic target identify problem. Experiment results show that the proposed method is more precise and less testing time cost than the traditional 1-a-1,1-a-r,SVM-DL,GADTSVM methods.