Empowering Extreme Communication: Propagation Characterization of a LoRa-Based Internet of Things Network Using Hybrid Machine Learning

The rapid growth of the Internet of Things (IoT) has led to the emergence of Low Power Wide Area Networks (LPWANs) to connect devices in a smart future world. However, challenges such as wireless channel propagation imperfections lead to signal power loss, higher retransmissions, and service quality degradation, ultimately limiting the efficiency of IoT implementation. This study investigates Long-Range (LoRa) IoT networks wireless channel propagation characterization in tropical environments using hybrid Machine Learning (ML) techniques. It specifically addresses the limitations of various established Path Loss (PL) models, such as Longley–Rice Irregular Terrain Model (ITM), through comprehensive evaluations across diverse scenarios, demonstrating their inadequacies in equatorial regions. Accordingly, four unique three-stage stacked ML-based semi-empirical PL models for LoRa communication were proposed. These models explore the combination of a Free Space PL (FSPL) with a Stepwise (STW) or linear Support Vector Machine (SVM) ML model, which is then integrated with an Ensemble of Bagged Trees (EBT) or Artificial Neural Network (ANN) ML model. With prediction accuracies reaching up to 89% for testing datasets, the FSPL-STW-EBT model showed the best performance, followed by the FSPL-SVM-EBT method. The latter two models showed the highest prediction accuracy for rural and suburban areas, 87% to 96%, across various datasets and categories. Additionally, these models achieved up to 95% accuracy in measurements taken on lake water surfaces. The study further emphasizes the balance between computational efficiency and predictive accuracy of these models, enhancing their suitability for real-time IoT applications in challenging environments. These findings highlight the potential of hybrid ML-based models to outperform conventional PL models and optimize IoT network design, planning, and deployment in tropical regions.