Residual-Based Analysis of the Mismatch Between LoRa Channel Models and Field Measurements in Urban and Rural Environments

Muhamad Bagus Fikril Alan (Department of Electrical Engineering, Faculty of Engineering, Universitas Negeri Surabaya, Surabaya, Indonesia)
Pradini Puspitaningayu (Department of Electrical Engineering, Faculty of Engineering, Universitas Negeri Surabaya, Surabaya, Indonesia)
Nurhayati Nurhayati (Department of Electrical Engineering, Faculty of Engineering, Universitas Negeri Surabaya, Surabaya, Indonesia)
Muhammad 'Aamir Nashrullah (Department of Electrical Engineering, Faculty of Engineering, Universitas Negeri Surabaya, Surabaya, Indonesia)
Nobuo Funabiki (Department of Electrical and Communication Engineering, Faculty of Engineering, Okayama University, Okayama, Japan)
Erwin Sutanto (Department of Physics, Faculty of Science and Technology, Universitas Airlangga, Surabaya, Indonesia)
Fahmi Fahmi (Department of Electrical Engineering, Faculty of Engineering, Universitas Sumatera Utara, Medan, Indonesia)

Eksakta : Berkala Ilmiah Bidang MIPA
Vol. 27 No. 03 (2026): Eksakta : Berkala Ilmiah Bidang MIPA (E-ISSN : 2549-7464) In Progress

This study investigates discrepancies between classical propagation models (Okumura-Hata, CI, FI) and LoRa field measurements in urban and rural environments. While rural areas exhibited linear path loss trends, urban scenarios showed distinct signal saturation, resulting in significant residuals even after model optimization. To address this, a residual-based analysis using Machine Learning (Linear Regression, Decision Tree, Random Forest) was proposed to map these systematic errors. The evaluation reveals that Random Forest (RF) significantly outperforms other algorithms, achieving an of 0.961 and an RMSE of 3.291 dB. These findings demonstrate that model mismatches follow deterministic patterns driven by environmental features rather than random noise. The study concludes that integrating ML-based residual compensation is essential for accurate radio planning in heterogeneous network deployments.