Comparative evaluation of linear and nonlinear regression models in predicting VLC channel response and BER performance

The current study explores the performance of an indoor Visible Light Communication (VLC) system using both linear and nonlinear regression methodologies. The analysis is conducted within the dimensions of a 5 m × 5 m × 3 m room, examining the channel frequency response, including magnitude and phase responses. Predictions are made using both linear and nonlinear regression models at distinct locations—center, side, and corner positions. Results indicate that the linear regression model outperforms the nonlinear approach, providing more accurate predictions of the true response. This conclusion is further supported by simulations of the bit error rate (BER) versus SNR for a 4 × 4 Multiple-Input Multiple-Output (MIMO) VLC system, employing both linear and nonlinear regression methods.