This study systematically investigated the complex factors influencing the mechanical properties of alkali-activated slag-fly ash (AASF) pastes and concrete through an integrated approach combining experimental investigations and machine learning techniques. The critical parameters governing compressive strength and elastic modulus were analyzed using random forest regression (RFR) and gradient boosting regression (GBR) models. The results show that the machine learning predictions demonstrate high accuracy, with deviations of compressive strength and elastic modulus predictions maintained within ±15% of the experimental values. Quantitative predictive formulas are established to enhance the efficiency and effectiveness of mechanical performance optimization. A dual-objective analysis framework reveals synergistic relationships between compressive strength and elastic modulus in both paste and concrete systems, providing effective pathways for mix proportion optimization. The results demonstrate a threshold effect in fly ash content: positive correlation with compressive strength at content below 25% (mass fraction), transitioning to negative correlation when ranging between 50% and 75% (mass fraction). This research presents an efficient intelligent solution for performance optimization of AASF materials while establishing a practical foundation for low-carbon material development in civil engineering applications.