Marine clay strengthened with recycled tiles: A novel hybrid model for evaluating unconfined compressive strength

As an eventual goal of various geotechnical experimental trials, a reliable assessment of the clay characteristics combined with recycled materials was put into consideration. Some key physical and mechanical parameters of marine clay (𝑀𝐶) amended with recycled tiles (𝑅𝑇) were used as independent variables in this study to estimate unconfined compressive strength (𝑈𝐶𝑆). Four novel hybridized methods were formed, combining the extreme gradient boosting (𝑋𝐺𝐵) model with various optimization algorithms such as the Black widow optimization algorithm (𝐵𝑊𝑂𝐴), Ant lion optimization (𝐴𝐿𝑂), Arithmetic optimization algorithm (𝐴𝑂𝐴), and Moth flame optimization (𝑀𝐹𝑂). In the 𝑋𝐺𝐵 model, optimization methods were utilized to determine the best appropriate value for determinative variables. Seven evaluation metrics were generated to measure the models' correctness. As well, uncertainty with a 95% confidence level (𝑈95) was applied by the 𝑡𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐 exam (𝑇𝑠𝑡𝑎𝑡). The findings show that all four models have a high level of accuracy in their 𝑈𝐶𝑆 prediction procedures, demonstrating the strong relationship between observations and predictions of the 𝑈𝐶𝑆 of mixed 𝑀𝐶, with 𝑅2 values for the learning and evaluating collections of minimum 0.9674 and 0.9888, correspondingly. With 𝑆𝐼𝑇𝑟𝑎𝑖𝑛=0.0253 and 𝑆𝐼𝑇𝑒𝑠𝑡=0.0264, the 𝑂𝐴−𝑋𝐺𝐵 system has the least scatter index (𝑆𝐼) values in comparison with others (Excellent performance). The 𝐴𝑂𝐴− 𝑋𝐺𝐵 model clearly has the least 𝑈95, indicating that it has a higher generalization potential. Ultimately, the hybridized 𝐴𝑂𝐴− 𝑋𝐺𝐵 concept may outperform others to be acknowledged as a suggested one from this research after evaluating the justifications of findings and comparisons with reviewed studies.