Optimizing 5G IoT Connectivity for Path Loss Reduction through Ant Colony Optimization and Support Vector Regression

Abstract

The study examines path loss models for IoT connectivity in 5G networks, emphasizing the importance of accurate estimation for coverage evaluation and base station installation while highlighting the limitations of empirical and deterministic models. Sophisticated modeling methods, including machine learning techniques and nature-inspired algorithms, are explored to address path loss issues in 5G IoT connectivity. Support Vector Regression (SVR) emerges as a promising approach, complemented by Ant Colony Optimization (ACO) to enhance model performance. The results of the evaluation show that the Free Space model has a relatively high mean squared error (MSE) of 6984.94, root mean squared error (RMSE) of 83.56, mean absolute error (MAE) of 81.18, and R-squared of - 18.07. This shows a less accurate prediction of path loss reduction. The log model also demonstrates poor performance with a high MSE of 15442.95, RMSE of 124.27, and MAE of 122.82. The negative R-squared is -41.15. The ITU-R model yields a lower MSE of 418.81, RMSE of 20.47, and MAE of 13.69, and the negative R-squared indicates a moderate fit. ITU-R model performs better than the Free Space and Log models but cannot accurately predict path loss reduction. Multiwall and Floor Model with MSE of 10063.42, RMSE of 100.32, MAE of 82.22, and a negative R-squared of -26.47. This exhibits higher errors. The SVR_KBF model produced a low MSE of 0.88, an RMSE of 0.94, an MAE of 0.27, and a high positive R-squared of 1.00. This shows a high level of accuracy in path loss prediction. An optimized SVR with an ACO algorithm outperforms all other models, with an extremely low MSE of 0.16, RMSE of 0.34, MAE of 0.05, and R-squared of 1.0. The optimized SVR model demonstrates superior path loss reduction for 5G IoT connectivity, showcasing its practical deployment potential in real-world scenarios.