Research On Network Traffic Modeling And Prediction Based On Alpha Stable Distribution

With the advent of the 5G era,network traffic has shown a skyrocketing trend.Network traffic modeling and predictive analysis have become the key to avoiding network congestion and improving network service quality.The conventional network traffic model did not consider the impulse,self-similarity,and volatility clustering of traffic data.Therefore,this study analyzed the self-similarity and volatility clustering of traffic based on the Alpha stable distribution theory,and improved The FARIMA model conducts network traffic modeling and forecasting research.Based on the analysis of long correlation,heavy-tail characteristics,volatility clustering and self-similarity theory,this thesis mainly conducts research from three aspects: First,in order to improve the accuracy of the autoregressive order and the moving average order in the FARIMA model,A network prediction model combining quantum genetic algorithm and FARIMA model is given;secondly,in order to reduce the difficulty of modeling and improve the degree of fitting,a FARIMA model improved based on the segmented bidirectional CUSUM algorithm is proposed;finally,for the FARIMA model Unable to accurately track and describe the problem of network traffic volatility,use the segmented two-way CUSUM algorithm to improve the FARIMA model combined with the FIGARCH model to track and describe the volatility,and propose an improved FARIMA-FIGARCH network traffic prediction model;The real traffic is used in the above three improved FARIMA models for simulation,and the simulation results show that the network prediction accuracy of the three improved FARIMA models has been significantly improved.The comparison of the prediction results of the above three improved FARIMA models shows that the improved FARIMA model based on the segmented two-way CUSUM algorithm and the improved FARIMA-FIGARCH model have smaller prediction errors,and the improved FARIMA-FIGARCH model has better peak prediction results.The results of the research can be effectively applied to network optimization and network design,which is of great significance to the development of network services.