Hierarchical Auto-Associative Polynomial Convolutional Neural Network With Gorilla Troops Optimization for an Effective Millimeter-Wave Path Loss Modeling in 5G-IoT Mobile Communication System

Abstract

Path loss modeling (PLM) for mmWave communications (mWC) is challenging due to the dynamic propagation environment, making it critical for effective 5G network planning and analysis. In this manuscript, hierarchical auto-associative polynomial convolutional neural network with gorilla troops optimization (GTO) is proposed for path loss modeling in mmWave 5G-IoT mobile communication systems (HA-PCNN-PLM-mWC-5G). The input images depicting buildings and roadways are obtained from Google Maps and enhanced local area multiscanning (E-LAMS) with distance dependencies are used to improve feature learning. Here, the improved bilateral texture filtering (IBTF) is applied to reduce noise in the input images and enhance image quality. Additionally, the spatial and spectral features are then extracted using the fast discrete curvelet transform with wrapping (FDCT-WRP), and these features are fed into the HA-PCNN model for path loss prediction. The model's performance is further improved through parameter optimization using GTO. Here, the implementation of the proposed model is done in Python tool and the performance metrics are analyzed. Thus, the proposed approach attains 6.2%, 2.27%, 4.08%, 11.88%, and 12.32% higher accuracy, 13.07%, 14.41%, 16.61%, 18.03%, and 9.08% lower mean squared error, and 27.55%, 24.05%, 23.48%, 20.05%, and 18.95% lower computation time than the existing approaches like AE-CNN-mmW-PLM-5G, ML-PLP-mWL-5G, CNN-mmW-PLM-FWA, ML-SRM-IoT, and PLP-EE-5G, respectively. Thus, the proposed model improves the accuracy of mmWave communication prediction in 5G-IoT systems, offering a more reliable solution for future 5G network implementations.