Alpha Stable Distribution Based FARIMA Modeling and Forecasting for Network Traffic Data

Abstract

Through research on network traffic in recent years, it is found that actual network traffic not only has long correlation but also has heavy-tailed characteristics. The Alpha stable distribution provides a very useful theoretical tool for such processes. According to the nature of the Alpha stable distribution, we need to improve the relevant parameter value method of the Fractional Autoregressive Integrated Moving Average model and improve the prediction accuracy of the model. In this paper, the Alpha stable distribution based Fractional Autoregressive Integrated Moving Average modelling method is introduced, and the validation of the method is analyzed using Fractional Gaussian Noise time series and impulse noise corrupted Fractional Gaussian Noise time series. Finally, the Alpha stable distribution based Fractional Autoregressive Integrated Moving Average model is applied to BC-p Aug89.TL data provided by Bell Labs. The analysis shows that the improved Fractional Autoregressive Integrated Moving Average model has a better prediction effect on actual network traffic and is a more representative research method.