Estimation of compressive strength of recycled aggregate concrete using advanced meta-heuristic algorithms and random forest analysis

This study uses random forest (RF) analysis to present a novel method for estimating the compressive strength (F_c) of concrete made with recycled aggregate (RAC). The intricate interactions between various components make it challenging to achieve the desired F_c through the best mixture proportions. The 441 samples are used with six inputs, including Coarse Aggregate to Cement Ratio, Cement Fineness, Fine Aggregate to Total Aggregate Ratio, Specific Gravity of Saturated Surface-Dry Aggregates, and Water Absorption of Aggregates. Algorithms for machine learning (ML) and artificial intelligence (AI) have shown significant promise for solving this complexity. In addition, a semi-empirical strategy is offered that seamlessly incorporates optimization methods to improve prediction precision. The accuracy of models is improved using the Improved Grey Wolf Optimizer (IGWO) and Flying Foxes Optimization (FFO). Therefore, three distinct models are produced: RFFO, RFIG, and a standalone RF model. Each of these models provides essential insights into accurate F_c prediction for RAC. Interestingly, the RFFO model outperforms all others with a remarkable R² value of 0.996 and a remarkably low RMSE value of 0.993. Furthermore, RF obtained a weak performance equal to 4.897 compared to other developed models, in which the RFFO achieved 0.986. These results demonstrate the RFFO model's skill in predicting RAC results while reiterating its accuracy and dependability. This technique has great potential for precise RAC prediction in the construction sector.