互联网传播行为的时序演化与预测

Timing evolution and prediction of Internet transmission behavior

互联网的传播行为对研究网络拓扑结构和动态行为的关系具有重要作用。选取CAIDA_Ark项目下不同地区4个监测点的有效路径样本数据,统计网络访问时间与访问直径,发现它们的相关性极弱,网络访问时间呈多峰重尾分布。采用非线性时间序列分析方法对网络访问时间演化序列混沌辨析,结果表明其时序演化具有混沌特征。在此基础上,引入Logistic方程建立网络传播行为预测模型,并用粒子群优化算法对模型参数取优,用4个监测点的网络访问时间序列对模型进行实验,从准确性和可用性这2个方面对模型进行评价,结果表明,短期内该模型能够对网络传播行为做出准确预测,在一段时期内,可作为网络行为演化预测的工具。

The transmission behavior of Internet plays an importance role in the research on the relationship between network topology structure and dynamic behavior. Selecting effective path samples in four monitoring points which from different regions authorized by CAIDA_Ark project and statistics network traveling time and traveling diameter, their correlation is very weak, network traveling time is presented on multi-peak and heavy tail distribution. Using nonlinear time sequences analysis method to identify the Chaos characteristics of network traveling time evolution sequences. The results show that their timing evolution has Chaos characteristics. Based on this, the Logistic equation was lead to establish network transmission behavior prediction model, and particle swarm optimization（PSO） was used to optimize model parameters. The model by the network traveling time sequences of four monitoring points was experimented, evaluated it from accuracy and availability, the results show that the model can predict network transmission behavior accurately in the short term. It can be used as a tool for predicting the network behaviors＇ evolution in a period of time.