Optimization machine learning models for selecting transmit antennas in 5G/6G systems

Abstract

Transmit antenna selection (TAS) plays a crucial role in improving the performance and spectral efficiency of 5G/6G systems. This study proposes to use the GridSearchCV method for hyperparameter optimization in two supervised learning models, support vector machine (SVM) and K-nearest neighbors (KNN), to optimally select antenna peers based on channel gain. These models were applied to Alamouti’s space-time block coding to improve performance, resulting in increased signal-to-noise ratio (SNR) and reduced bit error rate (BER). The results show that optimizing the hyperparameters led to a significant improvement in the performance of the SVM and KNN models. The SVM and KNN models were evaluated using a variety of metrics, with the SVM demonstrating superior predictive performance in terms of accuracy, average macro recall, average macro precision, average macro F1 score, and cross-validation score. Even before optimization, the SVM outperforms the KNN in terms of performance metrics. After optimization, this gap widens further, demonstrating the robustness of SVM for classification tasks. Although KNN is faster to train.