基于改进粒子群优化算法和CRNN的多类SVM分类

为了提高支持向量机(SVM)在多类分类中的分类效果,提出了一种基于改进粒子群优化(IMPSO)算法和协作式递归神经网络(CRNN)的多类SVM分类方法(IMPSO_CRNN_SVM算法).首先引入自适应惯性权重及自适应粒子变异,以此改进粒子群优化算法(PSO)在优化SVM参数过程中存在的容易陷入局部最优和早熟等问题; 然后基于多类SVM设计一个CRNN,并利用随机分配的训练集对该网络进行训练并构建最终决策函数,从而实现多类数据的“一次性”分类.最后利用3种数据集和实际应用对IMPSO_CRNN_SVM算法进行验证,结果表明IMPSO_CRNN_SVM算法的分类精度优于未进行参数优化的传统SVM算法、基本PSO 进行SVM参数优化的算法和未进行PSO参数优化的基于CRNN的多类支持向量机算法,因此IMPSO_CRNN_SVM算法具有一定的实用性.

Classification of multi-class support vector machines based on improved particle swarm optimization and CRNN

Aiming at the factors that affect the application of support vector machine(SVM)in multi-class classification, a multi-class SVM classification method(IMPSO_CRNN_SVM algorithm)based on improved particle swarm optimization algorithm(IMPSO)and cooperative recurrent neural network(CRNN)was proposed. Firstly, adaptive inertia weight and adaptive particle variation were introduced to improve the problem of local optimization and prematurity of particle swarm optimization algorithm(PSO)in the process of optimizing SVM parameters. Then, based on multi-class SVM technology, a CRNN was designed. The randomly assigned training set was used to train the network to construct the final decision function, so as to realize the "one -step" classification of multi-class data. Finally, the IMPSO_CRNN_SVM algorithm is verified by different data sets and practical applications. The results show that the classification accuracy of IMPSO_CRNN_SVM algorithm is better than that of SVM algorithm without parameter optimization or traditional PSO parameter optimization and multi-class SVM based on CRNN without parameter optimization, and it has certain practicability.