基于格兰杰因果关系贝叶斯网络的大规模无线局域网流量预测方法  

Predicting Large-Scale WLAN Traffic via Granger Causality Based Bayesian Network

作　　者：王浩[1] 吕云飞[1] 陈源宝[1] 彭云飞[1] 

机构地区：[1]武汉第二船舶设计研究所,武汉430064

出　　处：《电信科学》2015年第8期46-50,共5页Telecommunications Science

摘　　要：研究了大规模无线局域网内的流量特性,发现不同接入点间的流量存在格兰杰因果关系。流量的格兰杰因果关系说明,可以通过多个存在因果关系的接入点的历史流量,提高对目标接入点的当前流量预测的准确性。通过贝叶斯网络对存在因果关系的接入点流量进行建模,并利用多个接入点的历史流量对目标接入点的流量进行预测,提高了预测的准确性。最后,通过接入点数量大于100个的无线局域网的实际流量数据,验证了该方法的有效性及准确性,建立了一套完整的数据特征分析、建模及预测的流量数据处理流程。Granger causality existed between traffic at different access points of large-scale wireless LANs was discovered. The Granger causality illustrates that the historical traffic of access points that exist causality within target access points help predict the future of target access points with better accuracy than when considering information from the past of target access point alone. Bayesian network to model the causal relationship between access points and adopted a Gaussian mixture model （GMM） was used, as well as a weighted combination of several normal distribution functions in order to approximate the joint probability distribution in Bayesian networks. Finally, the traffic data in large-scale wireless LANs was imported, having hundreds of access points, to verify the accuracy of the proposed method, and a processing flow of analysis, modeling and prediction of traffic flow data was established.

关 键 词：无线局域网 流量预测 流量特性 格兰杰因果关系 贝叶斯网络 

分 类 号：TN925.93[电子电信—通信与信息系统]