Optimized precoding for massive MU-MIMO systems with KLDA dimension reduction and RNN-crossover GBO algorithm

Nowadays the communication of massive multi-user multiple-input multiple-output (MU-MIMO) takes an important role in wireless systems, as they facilitate the ultra-reliable transmission of data and high performance. In order to sustain massive user equipment (UE) with tremendous reliability and spectral efficiency, more antennas are deployed per base station (BS) in the MU-MIMO system. To overcome such problems, the recurrent neural network (RNN) with crossover-gradient based optimizer (GBO) model called RNN-crossover GBO is proposed for precoding in the MU-MIMO system. However, it is essential to diminish the complexity to attain the maximum sum rate for obtaining the optimal solution. Moreover, the kernel linear discriminant analysis (KLDA) dimensionality reduction technique is employed for mapping high dimensional data into a low dimension by considering a linear combination matrix. In order to obtain the best feature the GBO is employed that predict the optimal solution and restrict falling from the local solution. Furthermore, the crossover-GBO algorithm is applied with the RNN to estimate the output precoding matrix with considerable features to obtain the best search space. The experimental results revealed that the proposed method achieves higher performance with a higher sum rate and shows significant improvement in spectral efficiency (SE) values than the existing methods. SE rises due to the selection of closely associated features. This indicates the robustness and stability of the proposed model.