Machine learning-assisted optimization of the encapsulation of Helichrysum splendidum essential oil for food preservation applications

Abstract

• Chitosan-WPC matrices effectively encapsulated Helichrysum splendidum essential oil. • Machine learning models outperformed RSM in predicting encapsulation responses. • ANN-GA optimized formulations enhanced antioxidant and antifungal activities. • ML-GA provide robust strategy for optimizing functional food bioactive delivery. Postharvest losses pose a major challenge to food security, underscoring the need for natural, sustainable preservation strategies. Essential oils (EOs) are promising alternatives to synthetic preservatives, but their volatility, poor water solubility, and instability limit practical applications. This study optimized the encapsulation of Helichrysum splendidum EO using chitosan and whey protein concentrate (WPC) as wall materials. A central composite design under response surface methodology (RSM) was employed, with polymer ratio, EO concentration, and Tween-80 as factors, while water solubility, moisture content, antioxidant, and antifungal activities were evaluated as responses. In parallel, machine learning (ML) models, including support vector machines (SVM), Gaussian process regression (GPR), and artificial neural networks (ANN), were developed and compared with RSM for predictive performance. RSM analysis revealed that chitosan and EO concentration were the most influential factors on encapsulation outcomes, with significant interaction effects on functional properties. Among ML models, ANN and GPR demonstrated superior predictive accuracy compared to RSM, particularly for antioxidant and antifungal activity. The optimal formulation from the best performing model (ANN), coupled with a genetic algorithm, consists of 23.65% chitosan, 52.53% WPC, 22.75% EO, and 1.07% Tween 80. This formulation exhibited enhanced water solubility, reduced moisture content, and strong antioxidant and antifungal activities, confirming the complementary behaviour of the two polymers, where chitosan contributes antifungal activity and WPC enhances antioxidant potential. This study demonstrates the efficacy of integrating RSM with ML tools for encapsulation optimization and establishes H. splendidum EO in chitosan-WPC matrices as a promising natural preservative system for postharvest fruit preservation.