A decision-support framework for early-stage screening of sustainable and degradable plastics in a circular economy

• A framework to predict plastic degradation in soil and seawater was developed. • MCDA and machine learning were combined to improve prediction accuracy. • Decision tree models achieved high accuracy for both soil and marine cases. • Entropy-weighted scoring enhanced model robustness and stability. • A layered workflow links modeling with experimental and regulatory standards. Plastic pollution presents a significant and persistent threat to ecosystems. Degradable plastics are promising, yet their degradation varies widely depending on material properties and environmental conditions. Accurate prediction of environmental degradation is essential for advancing sustainable design and regulatory evaluation. We proposed a two-stage framework integrating MCDA with machine learning models to predict degradation rates in soil and seawater. Amongst these models, a depth-3 decision tree achieved the best accuracy. Sensitivity and robustness analyses highlighted the importance of molecular weight and functional-group features and confirmed the consistency of the entropy-weighted WSM configuration. Sensitivity and robustness analyses confirmed the effectiveness of Entropy weighting and WSM-based scoring in improving model stability and predictability. The framework enables early-stage, decision-relevant screening of diverse plastic materials. We outline a screening workflow that couples model outputs with experimental validation and standard protocols, supporting sustainable materials selection aligned with circular-economy goals.