粒子滤波最大熵算法在网络流量预测中的应用

Study about Network Traffic Prediction Base on Particle Filter Maximum Entropy Algorithm

研究网络流量预测优化问题,网络技术的发展使网络的流量增大。针对网络中对网络流量的不确定影响因素较多,同时由于传统的网络流量预测算法都是线性模型,无法适应网络流量非线性变化,从而导致预测精度不高等缺陷,提出了一种粒子滤波和最大熵算法原理相结合的新的网络流量预测模型。首先用模型捕捉原始数据的重构性,然后采用粒子滤波算法对最大熵进行优化处理,根据得到的预测结果作为约束的信息,采用优化后的最大熵得出预测的结果分布。最后采用算法与其它几种常见的流量预测算法相比较,仿真结果表明,改进方法比传统的几种网络流量预测算法具有更高的预测准确度和较高的泛化能力。

Optimization of network traffic prediction.Network,network traffic for the uncertain effects of many factors,and because of the traditional traffic prediction algorithms are based on linear model,nonlinear changes can not adapt to network traffic,resulting in higher prediction accuracy is not defective,is proposed based on examples of filtering and maximum entropy algorithm principle of combining the new network traffic prediction model.The first model to capture the raw data reconstruction,and then using the particle filter algorithm to optimize the maximum entropy,according to the prediction results of the information as a constraint,using the optimized maximum entropy algorithm to predict the results obtained distribution.Finally,the algorithm using real network traffic is predicted.The simulation results show that this method than the traditional network traffic prediction algorithm has several higher forecast accuracy and higher generalization ability.