<p>The environmental impact of conventional concrete necessitates sustainable alternatives incorporating industrial by-products and recycled materials. However, the complex interactions of these components often degrade performance in ways that black-box artificial intelligence models cannot mechanistically explain, limiting their practical utility for engineers. This study addresses this gap by developing interpretable stepwise regression models to predict the key properties of green concrete containing ground granulated blast-furnace slag (GGBFS), waste rubber powder (WRP), and recycled concrete aggregates (RCA). Utilizing a dataset of 204 experimental mixes, stepwise linear regression (SLR) and stepwise polynomial regression (SPR) frameworks were employed to derive explicit equations quantifying material interactions. Validated via 10-fold cross-validation, SPR models achieved superior accuracy (R² &gt; 99% for mechanical properties, 97.88% for chloride resistance) by incorporating quadratic and interaction terms. While the study’s predictive scope is currently limited to chloride ion penetration as the primary durability metric, the findings reveal key trade-offs: slag enhances both strength and durability, whereas WRP and RCA, though beneficial for waste reduction, degrade mechanical performance. Importantly, the SPR framework demonstrates that optimizing these ternary mixes—by prioritizing slag content and limiting WRP/RCA dosages—can enable the design of eco-efficient concrete with a potentially lower CO₂ footprint without compromising structural integrity. This work bridges the gap between black-box AI models and empirical methods by offering engineers actionable, transparent equations for sustainable concrete design.</p>