Prediction of Macrocell Traffic Congestion Using a Hybrid of Polynomial Neural Networks and Genetic Algorithms

Abstract

This work used a hybrid of Group Method of Data Handling (GMDH) neural networks and Genetic Algorithm (GA) to optimize busy hour traffic congestion prediction model in cellular macrocell. The optimized model was implemented in MATLAB environment using call setup success rate (CSSR) and busy hour (BH) traffic of the macrocell as input to the model and traffic channel (TCH) congestion of the macrocell as the target. The GA was used for the optimal layer selection pressure of GMDH neurons and on the average improved the prediction accuracy of GMDH model by reducing its mean absolute percentage error (MAPE) by 80%.