基于线性多尺度模型的计算机网络数据流量预测

为了解决网络安全监控问题,提出了一种用于预测网络流量的算法.通过多个不同尺度的线性模型进行网络数据的组合预测,每个尺度的线性模型由经过滤波器滤波后的部分原始数据估计得到,最终的预测流量数据由多个尺度线性模型的平均预测值得到.选择的线性模型为自回归滑动平均模型,且尺度较小的线性模型对应自回归滑动平均模型的阶数较高.结果表明,本算法的预测精度高,整体预测误差的均值在10-3量级.

Prediction of data flow in computer network based on linear multi-scale model

In order to solve the supervisory and control problems of network safety, an algorithm for the prediction of network flow data was proposed. The combined prediction of network data was carried out based on multiple linear models with different scales. The linear models with each scale were obtained through estimating the partial original data after filtering with a filter. The final predicted flow data were obtained from the average predicted values with multi-scale linear models. The selected linear models were the autoregressive moving average models. The linear model with a lower scale corresponds to a relative autoregressive moving average model with a higher order. The results show that the proposed algorithm has high predicted accuracy, and the mean value of entire prediction error is in the level of 10^-3.