Predicting Impact Strength of Natural Fiber Composites Using Optimized Gradient Boosting Approach

Natural fiber composites are gaining meteoric attention due to their eco-friendly nature and potential applications in various industries. Coir fiber, with India being its highest producer among tropical countries, holds immense significance due to its abundant availability and potential for sustainable applications. This research article presents an approach for predicting the impact strength of short coir-based composites (CBCs) using least-square gradient boosting algorithm (LSBoost). The dataset encompassed information on matrix type, fiber properties, and manufacturing process. The data analysis revealed the significant variation of impact strength with the varying fiber fraction and manufacturing process. Thereon, genetic algorithm (GA) was deployed in replacement to Bayesian optimization to optimize the hyperparameters of the LSBoost predictive model that may explore the hyperparameter space differently and potentially find superior configurations. The differently optimized LSBoost models were compared using standard performance metrics. The GA optimized LSBoost model exhibited better performance in comparison to Bayesian optimization but was computationally intensive. This study significantly contributes to the development of efficient and reliable techniques for predicting the impact strength of sustainable composites.