Estimation of frost durability of recycled aggregate concrete by hybridized Random Forests algorithms

An effective approach to promoting sustainability within the construction industry is the use of recycled aggregate concrete (𝑅𝐴𝐶) as a substitute for natural aggregates. Ensuring the frost resilience of 𝑅𝐴𝐶 technologies is crucial to facilitate their adoption in regions characterized by cold temperatures. The main aim of this study was to use the Random Forests (𝑅𝐹) approach to forecast the frost durability of 𝑅𝐴𝐶 in cold locations, with a focus on the durability factor (𝐷𝐹) value. Herein, three optimization algorithms named Sine-cosine optimization algorithm (𝑆𝐶𝐴), Black widow optimization algorithm (𝐵𝑊𝑂𝐴), and Equilibrium optimizer (𝐸𝑂) were considered for determing optimal values of 𝑅𝐹 hyperparameters. The findings show that all developed systems faithfully represented the 𝐷𝐹, with an 𝑅 2 for the train and test data phases of better than 0.9539 and 0.9777, respectively. In two assessment and learning stages, 𝐸𝑂 − 𝑅𝐹 is found to be superior than 𝐵𝑊𝑂𝐴 − 𝑅𝐹 and 𝑆𝐶𝐴 − 𝑅𝐹. The outperformed model's performance (𝐸𝑂 − 𝑅𝐹) was superior to that of 𝐴𝑁𝑁 (from literature) by raising the values of 𝑅 2 and reducing the 𝑅𝑀𝑆𝐸 values. Considering the justifications, as well as the comparisons from metrics and Taylor diagram's findings, it could be found out that, although other 𝑅𝐹 models were equally reliable in predicting the the frost durability of 𝑅𝐴𝐶 based on the durability factor (𝐷𝐹) value in cold climates, the developed 𝐸𝑂 − 𝑅𝐹 strategy excelled them all.