计算机应用与软件Issue(2)：167-170,4.DOI:10.3969/j.issn.1000-386x.2015.02.042
最优训练样本子集的 LSSVM 网络流量预测
FORECASTING NETWORK TRAFFIC BY LEAST SQUARE SUPPORT VECTOR MACHINE WITH OPTIMAL TRAINING SAMPLE SUBSET
摘要
Abstract
The selection of training sample has big influence on the generalisation ability of least square support vector machine (LSSVM).In order to improve the forecast accuracy of network traffic,we propose an LSSVM network traffic forecast model with optimal training sample subset.First we use density method to identify and eliminate the outliers in network traffic data as well as to eliminate the adverse effects of the outliers on fuzzy c-means (FCM)clustering results,and then employ FCM algorithm to cluster the processed network traffic data and select the optimal training set according to the minimum distance between input vector of forecasting point and clustering centre,strengthen the regularity of training set,and lessen the dependence of LSSVMon training set as well.Finally,we use LSSVMwhich has good nonlinear forecasting ability to learn the training set and to build network traffic forecast model,and through simulation experiment to test the performance of the model.Simulation results show that compared with other models,the proposed one improves the forecast accuracy of network traffic and speeds up model’s training,so the forecast results are more stable and reliable.关键词
网络流量/最小二乘支持向量机/模糊均值聚类/密度方法/预测精度Key words
Network traffic/Least squares support vector machine/Fuzzy c-means clustering/Density method/Forecast accuracy分类
信息技术与安全科学引用本文复制引用
张运涛,邢晨..最优训练样本子集的 LSSVM 网络流量预测[J].计算机应用与软件,2015,(2):167-170,4.基金项目
浙江省教育厅项目 （）
