Comparison of Artificial Neural Network and Response Surface Methodology for Predictive Modelling of Dielectric Properties of Pineapple Leaf Fibre Reinforced Epoxy-Based Composites

This paper explores the capabilities of Artificial Neural Network and Response Surface Methodology implemented on the topic of dielectric properties of polymer composites subjected to varying loading rates of natural fibres. For this paper, the dielectric properties of pineapple leaf fiber reinforced epoxy-based composites were studied based on its dielectric constant, loss factor and dissipation factor, where the factors were predicted using Artificial Neural Networks and Response Surface Methodology. Artificial Neural Network was carried out using Sckit-Learn and MATLAB 2021b and a comparison between both was carried out. Response Surface Methodology was performed using MINITAB software. The accuracy of Artificial Neural Network and Response Surface Methodology were discussed. Artificial neural network implemented using Sckit-learn machine learning when compared to MATLAB R2021b model was observed to produced results which have higher coefficient of determination ( \({R}^{2}\) ) and mean-squared error (MSE) which were lower. The comparison of artificial neural network and response surface concluded with a comparison of results obtained by MATLAB R2021b against MINITAB. The comparison resulted with artificial neural network being superior compared to response surface methodology as supported by the higher coefficient of determination ( \({R}^{2}\) ) and lower mean-squared error (MSE). A comparison between the predicted values of artificial neural network against the experimental data and the predicted values of response surface methodology against the experimental data supported the conclusion of this research.