An Optimization Model for Traffic Distribution Forecasting in Packet-Switching Networks

Traffic distribution forecasting is an essential step in network planning for packet-switching networks. It is frequently necessary to forecast and develop an optimal network configuration to meet the requirements of new traffic demands or changes in the existing demands. Several methods have been developed to forecast network configurations. While these studies have revealed some interesting traffic characteristics, little progress has been made in developing good models for the purpose of traffic engineering and performance prediction. We propose a novel multi-objective optimization model for traffic distribution forecasting in packet-switching networks by mapping these networks into multi-commodity networks. Initially, the radial basis function (RBF) network is used to monitor and learn the current real-traffic distribution. Next, a quadratic model is used to calibrate these functions for a precise traffic distribution. The proposed multi-objective optimization method can effectively and efficiently forecast the traffic distribution of packet-switching networks in both crisp and fuzzy environments. A numerical example is presented to demonstrate the application and effectiveness of this model