Recycled Aggregate Concrete (RAC) represents a significant innovation aimed at reducing the carbon footprint in the construction industry. Over the past few decades, numerous investigations and experiments have confirmed the viability of RAC as a construction material when the optimal combination of recycled and natural aggregates is used. This study seeks to further enhance the application of RAC by providing a robust framework for determining the optimal RAC mixture. To achieve this, machine learning is developed to predict the compressive strength of RAC by considering various mixture properties. To improve the accuracy of these predictions, the Symbiotic Organism Search (SOS) metaheuristic algorithm is employed, not only to fine-tune the machine learning hyperparameters but also to select the most suitable model. In this study, the SOS algorithm is tasked with choosing between Artificial Neural Networks (ANN), Support Vector Machines (SVM), or Random Forests (RF), based on predefined upper and lower bounds for their hyperparameters. The resulting machine learning model is then integrated with the novel Pareto Front Symbiotic Organism Search (PF-SOS) to generate a Pareto front of optimal mixtures, with compressive strength and production cost as the objectives. To validate the efficiency of the proposed method, the PF-SOS results are compared with those from other well-known multi-objective optimization algorithms. The findings demonstrate that PF-SOS offers faster convergence and a broader range of mixture options within the same function evaluation limit.