多子种群PSO优化SVM的网络流量预测
发布时间：2018-12-25 07:41
【摘要】：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 Optimization 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 Swarm Optimization,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.
[Abstract]:In view of the time-varying and non-stationary characteristics of network traffic, in order to improve the accuracy of network traffic prediction, a particle swarm optimization (PSO) algorithm and a support vector machine (Multi-Subpopulation Particle Swarm Optimization and Support Vector Machine,) network traffic prediction model based on "multi-subpopulation" mechanism are proposed. MSPSO-SVM) Firstly, support vector machine (Support Vector Machine,SVM) parameters are encoded into particle position strings, and the minimum cross-validation error of network training set is used as the parameter optimization objective, and then the optimal SVM parameters are found through the information exchange between particles. The "multi-sub-population" mechanism is introduced to solve the premature stagnation defect of particle swarm optimization (Particle Swarm Optimization,PSO) algorithm. Finally, the network traffic prediction model is established according to the optimal parameters, and the actual network traffic data are used for simulation test. The results show that compared with other prediction models, MSPSO-SVM can obtain better SVM parameters and improve the precision of network traffic prediction, which is more suitable for complex and changeable network traffic prediction.
【作者单位】： 华东交通大学信息工程学院;
【基金】：江西省教育厅科学技术研究项目资助(GJJ12686)
【分类号】：TP393.06
[Abstract]:In view of the time-varying and non-stationary characteristics of network traffic, in order to improve the accuracy of network traffic prediction, a particle swarm optimization (PSO) algorithm and a support vector machine (Multi-Subpopulation Particle Swarm Optimization and Support Vector Machine,) network traffic prediction model based on "multi-subpopulation" mechanism are proposed. MSPSO-SVM) Firstly, support vector machine (Support Vector Machine,SVM) parameters are encoded into particle position strings, and the minimum cross-validation error of network training set is used as the parameter optimization objective, and then the optimal SVM parameters are found through the information exchange between particles. The "multi-sub-population" mechanism is introduced to solve the premature stagnation defect of particle swarm optimization (Particle Swarm Optimization,PSO) algorithm. Finally, the network traffic prediction model is established according to the optimal parameters, and the actual network traffic data are used for simulation test. The results show that compared with other prediction models, MSPSO-SVM can obtain better SVM parameters and improve the precision of network traffic prediction, which is more suitable for complex and changeable network traffic prediction.
【作者单位】： 华东交通大学信息工程学院;
【基金】：江西省教育厅科学技术研究项目资助(GJJ12686)
【分类号】：TP393.06
【参考文献】
相关期刊论文 前6条
1 姜明;吴春明;张e，
本文编号：2390871
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2390871.html
