A Greedy Algorithm-Based BP Neural Network for User Association in RetNets

User association is one of the vital issues for Heterogeneous Networks (Hetnets) to give prominence to its advantages. For the purpose of finding desired solution, we first formulate user association in HetNets as an optimal problem with the object that balance the network throughput and traffic load on Macro Base Stations (MBSs) and Small BSs (SBSs). A greedy algorithm to solve the problem is given at the first place in this paper. However, computation and time cost are too heavy to implement it in practice. For this reason, we propose a Greedy Algorithm-based Back Propagation Neural Network (GA-BPNN) algorithm to reduce complexity while approximating performance. In addition, clusters of BSs that are suitable to implement Coordinated Multi-Point (CoMP) technique to cell-edged users can be selected simultaneously according to the output of the proposed GA-BPNN. Simulation proves that the proposed GA-BPNN method closely achieves the optimal solution obtained by the greedy algorithm but time cost is significantly shortened.