A Large Language Model-based Generative Artificial Intelligence and hybrid machine learning framework for prediction and inverse design of jute fiber concrete

The construction industry requires sustainable alternatives to conventional concrete to reduce environmental impacts while maintaining structural performance. Jute fiber concrete (JFC) is a promising low-carbon material; however, optimizing its mechanical properties is challenging because of complex nonlinear interactions among fiber characteristics and mix-design parameters. This study proposes an integrated framework combining Generative Artificial Intelligence (GEN-AI) with Decision Tree Boosting (DT-Boosting), Gaussian Process Regression (GPR), and GPR enhanced with Bayesian Optimization (GPR-BO) to predict and optimize JFC mechanical properties. An experimental dataset containing six input variables were expanded using a Large Language Model-based GEN-AI approach and validated through distribution analysis, boxplot comparisons, and Kendall correlation analysis. GPR-based models significantly outperformed DT-Boosting, reducing prediction errors by approximately 40–50%. The GPR-BO achieved the most consistent performance, with testing R 2 values of 0.9242 for compressive strength and 0.9682 for tensile strength. However, the SHAP and Partial Dependence Plot analyses identified jute fiber volume fraction and fiber content as the most influential parameters, while the water–cement ratio exhibited a strong negative effect. Probabilistic validation confirmed accurate reproduction of strength distributions. Furthermore, integrating GPR-BO with NSGA-II enabled inverse mix design, automatically generating optimized JFC mixtures satisfying user-defined mechanical performance requirements. The optimization employed six mix-design decision variables and minimized both normalized prediction error and fiber volume fraction to identify Pareto-optimal solutions. The resulting framework extends forward prediction to practical inverse design, reducing trial-and-error experimentation and material waste while supporting resource-efficient, performance-based, and sustainable concrete mix development for engineering applications.