An Internet Traffic Forecasting Model Adopting Radical Based on Function Neural Network Optimized by Genetic Algorithm

摘要

Traditional traffic forecasting model is hard to show non-liner characteristic of Internet. Neural networks and genetic algorithm are representatives of modern algorithms. Considering that BP neural networks model is easy to take local convergence, this paper put forward genetic algorithm optimizing weight and bias value of Radial Based Function Network (GA-RBF), made a Internet traffic forecasting model which is relative with p steps and ahead of l steps, overcame the limitations of traditional forecasting algorithm model and BP neural networks algorithm. To prove the effectiveness and rationality of this algorithm, we forecasted the China Education Network Main Port traffic with GA-RBF neural networks. According to the analysis, we find that the GA-RBF forecasting effect is obviously better than BP neural networks. The conclusion shows that it is one of available and effective ways to use GA-RBF artificial neural networks to do Internet traffic forecast.