Optimizing Path Loss Prediction: Advancements in Signal Parameter Estimation through Wavelet Neural Networks

Abstract

This study assesses the effectiveness of Neural Net- works (NN) and Wavelet Neural Networks (WNN) in predicting signal strength in wireless communication systems. WNNs, which integrate wavelet theory with NN architecture, demonstrate supe- rior performance. Significant reductions in Mean Squared Error (MSE) from 32.93 (NN) to 11.94 indicate improved precision for WNNs. Root Mean Squared Error (RMSE) decreases from
5.73 (NN) to 3.45 with WNNs, highlighting more consistent predictions. Mean Absolute Error (MAE) decreases from 4.62 (NN) to 2.75, showcasing enhanced accuracy in WNNs.
For predicting Average Distance, WNNs outperform NNs with lower MSE (59.4 vs. 106.65), RMSE (7.7 vs. 10.32), and MAE (5.73 vs. 7.878). The study, conducted on the Colab platform using Python, emphasizes that incorporating wavelet transforms enhances the model’s ability to recognize complex signal propagation patterns.