NEURAL NETWORK MODEL DEVELOPMENT FOR PATH LOSS PREDICTION IN EVOLVING COMMUNICATION TECHNOLOGIES

ABSTRACT

The paper presents a proposed data-driven path loss prediction over conventional models for wireless communication design, especially in the setting of dynamic signal variations in advanced Fifth Generation (5G) technologies. The research analyzes key parameters to train a multi-layered neural network using a robust data-driven model based on neural network architecture. Significant correlations show that environmental factors and path parameters affect path loss. The proposed three-layer model, trained over 150 epochs, outperforms traditional models in Mean Absolute Error, Mean Squared Error, and Root Mean Squared Error. With accuracy measures, it outperforms Okumura-Hata, Costa 231, and Egli models. The study uses TensorFlow and Python to explore 5G wireless communication parameter-path loss interactions for optimum network design and resource allocation. The simulation-based neural network path loss prediction method improves accuracy and reduces estimation mistakes. Future research includes enhanced machine learning techniques and 5G's impact on path loss prediction.

Keywords: Path loss, Prediction, Wireless Communication, Simulation, Neural Network Architecture, Fifth Generation (5G) technologies