ABSTRACT: In this paper, an improved hybrid model is proposed for signal loss prediction in Urban Macro (UMa) environments. The model combines the strengths of Long Short-Term Memory (LSTM) networks and exponential functions (EXP) to address limitations in both approaches. The exponential function captures the general trend of signal loss in UMa environments, while the LSTM network learns complex underlying temporal patterns in the data. The proposed hybrid model is evaluated using real-world data collected from drive tests in Port Harcourt, Nigeria. The results show significant improvements over the standalone EXP and LSTM models, achieving a 48% reduction in Mean Squared Error (MSE) compared to the EXP model and a 73% reduction compared to the LSTM model on unseen testing data. This underscores the strength and superiority of the hybrid approach over conventional signal loss models in UMa environments.
Keywords: LSTM, Signal, Prediction, Machine Learning, Wireless, Communication.
[1] C. Haslett, Essentials Of Radio Wave Propagation. Uk: Cambridge University Press, 2008.
[2] Huawei, "Network Planning Guide," Huawei Technol. Co., Ltd, No. 17, 2024, [Online]. Available: Https://E.Huawei.Com
[3] K. T. Herring, J. W. Holloway, D. H. Staelin, L. Fellow, And D. W. Bliss, "Path-Loss Characteristics Of Urban Wireless Channels," Ieee Ttransaction Antennas Propag., Vol. 58, No. 1, Pp. 171–177, 2010.
[4] Y. Zhang, J. Wen, G. Yang, Z. He, And J. Wang, "Path Loss Prediction Based On Machine Learning : Principle , Method , And Data Expansion," Mdpi Appl Sci, 2019, Doi: 10.3390/App9091908.
[5] I. Risi, C. Ogbonda, J. Isabona, S. E. Olukanni, And P. Harcourt, "A Wavelet Genetic Algorithm Based Signal Path Loss Model For Evolved Nodeb Cell Sites Planning And Effective Coverage Of Wireless Communication System," Int. J. Innov. Eng. Sci. Technol., Vol. 12, No. 2, 2023.
