Traffic Prediction in Telecommunications Networks: A Combined Forecast Method Based on Adaptive Genetic Algorithm

Combined forecast method is an important research direction in forecast field. The relevant research has shown many advantages of combined forecast. However, how to acquire efficiently the weight coefficients of combined forecast method to get the predicted value with minimum error is often hard to solve efficiently. Because the adaptive parameter real-coded genetic algorithm (APRGA) not only has the properties such as the global optimization, parallelism and strong stability etc., but also has shown the excellent performance of faster convergence and more chance of getting global optimal solution than both the real-coded genetic algorithm (RGA) and simple genetic algorithm (SGA). In this paper, APRGA is, therefore, introduced to solve the weight coefficients of combined forecast method. To compare the combined forecast method based on APRGA with the other forecast method, the time-series of telephone traffic in mobile communication networks are introduced. The experimental results show that combined forecast method based on APRGA has the excellent performance of higher prediction precision, faster convergence etc..