The present study focuses on developing a machine learning model to design concrete compositions using rubber aggregates. The Gradient Boosting (GB) model predicts compressive strength and modulus of elasticity with high accuracy on the control dataset, having a coefficient of determination R2 = 0,9983 and a root mean square error RMSE = 0,8947 MPa in predicting compressive strength and R2 = 0,9971 and RMSE = 0,7473 GPa in predicting elastic modulus. Global and absolute SHAP values are used to evaluate the influence of the eight input variables on compressive strength and modulus of elasticity. The influence of factors on compressive strength is arranged in the order: Crumb rubber (CR) > Tire rubber (TR) > Cement (C) > Water (W) > Silica fume (SF) > Coarse aggregate (CA) > Sand (S) and with the elastic modulus in the order CR > TR > W > C > CA > S > SF. Water content, TR, and CR are to reduce the strength and elastic modulus of rubber concrete with increasing content, while the remaining factors are beneficial for improving compressive strength and elastic modulus when increasing content.