Estimating path loss in mobile communication systems using ensemble learning approaches in 3D geographic information system model

Perihanoglu G. M., Karaman H., Akgül T.

Physical Communication, cilt.77, 2026 (SCI-Expanded, Scopus)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 77
  • Basım Tarihi: 2026
  • Doi Numarası: 10.1016/j.phycom.2026.103189
  • Dergi Adı: Physical Communication
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC
  • Anahtar Kelimeler: Communication systems, Ensemble boosting models, Geographic information system (GIS), Propagation path loss model (PL)
  • Van Yüzüncü Yıl Üniversitesi Adresli: Evet

Özet

Accurate path loss estimation is critical for the planning and optimization of mobile communication networks, particularly in urban environments where terrain and built structures significantly impact signal propagation. This paper presents an integrated approach utilizing Geographic Information Systems (GIS) and ensemble machine learning models to enhance path loss prediction. The contributions of this study are as follows: (I) A GIS-based framework is developed to incorporate Digital Twin (DT) of the study area with 3D geographic features, including terrain elevation and building heights, into empirical and machine learning-based path loss models. (II) The study evaluates four empirical models–Free Space, COST-231, Ericsson 9999, and SUI–alongside three ensemble learning models–Gradient Boosting Machine (GBM), Extreme Gradient Boosting (XGBoost), and Categorical Boosting (CatBoost). Comparative analysis is conducted using real-world signal measurements at 900 MHz, 1800 MHz, and 2100 MHz frequencies. Results indicate that ensemble learning models outperform traditional empirical models, with CatBoost achieving the highest prediction accuracy. The integration of 3D spatial data and ensemble learning algorithms enables more precise estimation of signal attenuation, demonstrating significant improvements in mobile network design. (III) Finally, this study highlights the potential of GIS-enhanced machine learning approaches for future network deployment, including applications in DT, 6G and beyond.