Abstract—This research paper presented an adaptive hybrid system predictor that employed an adaptive Linear Mean Square (LMS) filtering technique for dataset de-noising, passed through a combined Adaptive Linear Element (ADALINE) and Multi-Layer Perceptron Neural Network (MLP) for improved neural network training and prediction over a conventional MLP model. The neural network training performances of the designed filtered adaptive hybrid system predictor showed improved training and prediction results over the conventional MLP using measured data from Long-Term Evolution (LTE) microcell environment from a Line-of-Sight (LOS), termed as location- 1, and a Non-Line-of-Sight (NLOS), termed as location-2. The neural network models' training results analysis was carried out using 1st order statistical performance indices, the coefficient of Regression (R), the Root Mean Square Error (RMSE), the Mean Squared Error (MSE), the Standard Deviation (SD), and the Mean Absolute Error (MAE). The statistical performance indicators measured the closeness of the prediction values to the measured values during neural network training, using two training algorithms, the Levenberg-Marquardt (LM) and the Bayesian Regularization (BR) training algorithms. The output of the designed adaptive hybrid system predictor showed superior and optimal prediction of the measured dataset over the conventional MLP for both the LOS location-1 and the NLOS location-2. The prediction results also demonstrated better prediction of the neural network models using the BR training algorithm over the LM training algorithm. However, this comes with a certain tradeoff, such as increased training time for the BR training algorithm.

Keywords—dataset de-noising, least mean square filter, adaptive linear element, multi-layer perceptron neural network model, 1st-order statistical performance indices, training algorithms, signal power loss

Cite: Virginia C. Ebhota and Thokozani Shongwe, “Adaptive Hybrid Neural Network Predictor over a Multi-layer Perceptron Neural Network Model for Improved Electromagnetic Signal Power Loss Prediction," Journal of Communications, vol. 21, no. 2, pp. 245-260, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).