Load Balancing using Backpropagation in Mobile Networks

Mobile traffic is increasing rapidly as the number of mobile subscribers are increasing everyday. This results in uneven traffic in certain areas. Load balancing is a complex problem with dynamically varying inputs and outputs. This complex system is difficult to be modeled with conventional mathematical tools/methods.We use a multilayer recurrent neural network to model this problem. Backpropagation algorithm is used to train this multilayer recurrent neural network. Initially, the network is trained with previous mobile traffic data and tested for different traffic conditions.It is found to predict the loads in the base stations. This information helps in balancing the load on the base station by borrowing channels from an appropriate base station.