Improving millimetre-wave path loss estimation using automated hyperparameter-tuned stacking ensemble regression machine learning

• DOI: 10.1016/j.rineng.2024.102289
• Publication Date: 2024-05-20T16:36:52Z
• Authors: Johnson O. Afape; Alexander A. Will...; Modupe E. Sanyaolu; Obiseye O. Obiyemi; Katleho Moloi; Janet O. Jooda; Oluropo F. Dairo
• Keywords: Hyperparameter; Millimetre wave; Ensemble Learning; Hyperparameter Optimization

Path loss prediction is a crucial aspect of designing and operating wireless communication systems, especially in the millimetre-waves (mmWaves) frequency bands. However, these bands are associated with climate-related challenges: rain attenuation, free space path loss. To address challenges, an advanced stacking ensemble-regression machine learning (SEML) model automated hyperparameter tuning (AHT) was proposed. The AHT-SEML leverages multiple base regressors integrated meta-regressor. model's performance optimised using AHT technique. efficiency tested simulated data from Composite 3D Raytracing-Image-Method propagation across four sub-Saharan cities, at mmWaves frequencies. compared to three empirical models, namely Close-In (CI), Floating Intercept (FI), Alpha-Beta-Gamma (ABG), evaluation metrics such as Mean Absolute Error (MAE) Root Squared (RMSE). outperformed other models cities all frequencies scenarios highest Index Agreement lowest Bayesian information criterion. Model confidence set (MCS) analysis CI benchmark indicates that except performed below critical t-value 2.3530 95% level degree freedom 3, implying no significant differences their MAEs CI. AHT-SEML's t-statistic values exceed this t-value, indicating statistically better than models. Similarly, F-statistics 29.45 26.54 correspond p-values for MAE RMSE, respectively, corroborating performance.