Improving 6G Network Performance with Linear Regression and Stacking GBM

The constant increase in demand for faster data transmission in networks is anticipated to lead to significant challenges, such as congestion and performance degradation. Therefore, proactive investigation of network performance is vital to enhance both operational efficiency and the user experience. This study explores the application of machine learning (ML) techniques to optimize the performance of sixth-generation (6 G) wireless networks. Specifically, we propose a novel stacking-based ensemble learning approach that leverages a gradient boosting machine (GBM) as the meta-model with a linear regression (LR) based imputation method to accurately predict network throughput. The experimental results demonstrate the model's strong predictive accuracy, achieving an R2 of 0.8482, a root mean square error (RMSE) of 0.0948, and a mean absolute error (MAE) of 0.0600. These findings underline the model's effectiveness in capturing dynamic network behavior, marking an important step toward the development of intelligent, adaptive 6 G communication systems.