基于DWT的多模型组合社交网站访问量预测  

Based on DWT Multi-Model Combined Prediction of Page Views on the Social Networking Site

作　　者：王少帅 宋礼鹏[1] Wang Shaoshuai;Song Lipeng(Data Science and Technology,North University of China,Taiyuan 030051)

机构地区：[1]中北大学大数据学院,太原030051

出　　处：《信息安全研究》2018年第4期364-368,共5页Journal of Information Security Research

基　　金：国家自然科学基金项目(61379125)

摘　　要：针对局域网内社交网站访问量变化的不确定性而导致其预测精度低的难题,提出一种基于离散小波变换(DWT)的多模型组合预测模型.该模型利用DWT将局域网内社交网站访问量时间序列分解成反映序列总体变化规律的周期分量与体现了序列细节性变化规律的残余分量2部分,并分别使用高斯过程回归模型(GPR)和加权近邻模型(WNN)进行针对性预测.通过收集中北大学局域网内各大主流社交网站访问量数据对模型进行实验仿真.结果表明,相对于其他模型,提出模型的预测精度有了进一步提升.Because of the uncertainty of the change of the page views on the social networking site in local area network,in order to solve the problem of low prediction accuracy,we proposed a prediction model which combined multiple models based on discrete wavelet transformation (DWT).The model uses DWT to decompose the time series of social networking sites in local area network into two parts,one is the periodic components reflecting the general variation laws of the series and the other is the residual components reflecting the detail variation laws of the series, then use the Gaussian process regression (GPR)and weighted nearest neighbors (WNN) separately to targeted predict.Through the collection of the data of the page views on the major social networking sites in local area network of the North University of China to experimental simulation.The result shows,compared with other models,the prediction accuracy of our model has further improved.

关 键 词：离散小波变换 周期分量 残余分量 高斯过程回归 加权近邻 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]