多子种群PSO优化SVM的网络流量预测

针对网络流量的时变性和非平稳性特点,为提高网络流量预测精度,提出一种“多子种群”机制的粒子群算法和支持向量机的网络流量预测模型（Multi-Subpopulation Particle Swarm Opti-mization and Support Vector Machine,MSPSO-SVM）.首先支持向量机（Support Vector Machine,SVM）参数编码成粒子位置串,并根据网络训练集的交叉验证误差最小作为参数优化目标,然后通过粒子间信息交流找到最优SVM参数,并引入“多子种群”机制,解决粒子群优化（Particle SwarmOptimization,PSO）算法的早熟停滞缺陷,最后根据最优参数建立网络流量预测模型,并采用实际网络流量数据进行仿真测试.结果表明,相对于其他预测模型,MSPSO-SVM可以获得更优的SVM参数,网络流量预测精度得以提高,更加适用于复杂多变的网络流量预测.

Network traffic prediction based on SVM optimized by multi-subpopulation particle swarm optimization algorithm

The network traffic has time-varying and nonlinear characteristics, in order to improve the prediction accuracy of network traffic, a network traffic prediction model based on multi-subpopulation particle swarm optimization algorithm and support vector machine （MSPSO-SVM） is proposed in this paper. Firstly, SVM parameter is encoded into the position of the particle, and minimum the cross validation error of network training set is taken as optimal target. Then the parameters of SVM are obtained by the exchange information among particles, and multi-subpopulation is introduced to keep the diversity of particle swarm. Finally, network traffic prediction model is built according to the opti- mum parameters, and the simulation test is carried out on actual traffic data network. The results show that, compared with other prediction models, the proposed model can get the better parameters, and network traffic prediction accuracy can be improved. It is more suitable for complex network traf- fic prediction.