Abstract: Active ingredients in functional foods often fail to exert their intended efficacy due to poor stability, low solubility, inadequate bioavailability, and limited health effects. Shelf-stable and efficacy-enhanced delivery systems (SSEEDS) have emerged as a pivotal strategy to address these challenges by enabling precise delivery with high loading capacity, stability, and potency through various approaches, including dispersion and solubilization, stabilization and encapsulation, targeted release control, absorption enhancement, and synergistic formulation. However, traditional construction methods, relying on empirical trial-and-error, suffer from low efficiency and poor predictability. This review summarizes recent advances in the application of big data and machine learning (ML) for the intelligent construction of SSEEDS. It systematically explores their roles in functional component screening, carrier structure design, release behavior prediction, and multi-objective process optimization. Special emphasis is placed on case studies involving ML modeling for SSEEDS, prediction of release kinetics, and process regulation via Bayesian optimization. The advantages of ML in improving encapsulation efficiency, prolonging stability, and enhancing bioaccessibility are elucidated. Finally, this paper identifies prevailing challenges including data fragmentation, limited model generalizability, empirical dependence, and the complexity of cross-scale coupling, it also proposes integrating federated learning, transfer learning with few-shot enhancement, explainable AI, and digital twin technologies to address these challenges. This review aims to provide valuable technical insights and methodological guidance for the intelligent construction of SSEEDS for functional foods.

Key words: functional foods; shelf-stable and efficacy-enhanced delivery systems; big data; machine learning; release behavior prediction; Bayesian optimization; intelligent design