Machine Learning Prediction of Mechanical Properties of Nano Fly Ash-Modified Natural Fibre-Reinforced Composites

In this study, machine learning (ML) algorithms viz., Random Forest, eXtreme Gradient Boosting (XGBoost), and Least Absolute Shrinkage and Selection Operator (LASSO) were employed to predict the hardness, tensile, flexural, and impact strengths of nano fly ash-reinforced natural fibre composites. Three types of natural fibres (abaca, hemp, and kenaf) and varying nano fly ash content (1, 3, 5 vol.%) were attempted. The ML models, developed in Python environment, are trained using 80% of the data and the remaining were used for validation. Among the attempted models, Random Forest model achieved the highest prediction accuracy (99%) along with the lowest RMSE and SMAPE values. The XGBoost model followed closely with a 97% accuracy. In contrary, the accuracy of the LASSO varies between 45% and 75% for various mechanical properties along with higher RMSE and SMAPE values. Furthermore, a graphical user interface (GUI) was developed to facilitate data input, model selection, and visualization of predicted results.