基于遗传优化加权相关向量机的网络热点预测模型  

NETWORK HOTSPOT EVENTS PREDICTION MODEL BASED ON GA-WRVM

作　　者：李睿[1] 赵保华[2] 王朋[1] 

机构地区：[1]阿坝师范高等专科学校网络管理中心,四川汶川623002 [2]阿坝师范高等专科学校图书馆,四川汶川623002

出　　处：《计算机应用与软件》2015年第6期312-315,319,共5页Computer Applications and Software

基　　金：阿坝师专青年基金项目(ASC1224);四川省教育厅重点课题(13ZA0038)

摘　　要：为了提高网络热点事件的预测精度,提出一种基于遗传优化加权相关向量机的网络热点事件预测模型(GA-WRVM)。首先对每个样本数据的噪声方差加上一个权值系数得到加权相关向量机,然后用组合核函数代替相关向量机的单一核函数,并采用遗传算法对核参数进行优化建立预测模型,最后利用网络热点数据进行仿真实验分析。结果表明,相对于神经网络和支持向量机,GA-WRVM降低了网络热点事件的预测误差,获得了更高的网络热点预测精度,具有更强的泛化能力。In order to improve the prediction accuracy of network hotspot events,in this paper we propose a new network hotspots events prediction model,which is based on using genetic algorithm to optimise weighted relevance vector machine （GA-WRVM）.First,we impose the weighting coefficient to the noise variance of every sample data to get the weighted relevance vector machine （WRVM）.Then we use the combined kernel function to take place the single kernel function of WRVM,and employ genetic algorithm to optimise kernel function parame-ters to build the prediction model.Finally,we carry out the simulation experiment by using network hotspots event data.Results show that compared with the neural network and support vector machine,GA-WRVM reduces the prediction errors of network hotspot events,and obtains higher prediction accuracy,thus has stronger generalisation ability.

关 键 词：热点事件 组合核函数 加权相关向量机 遗传算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]