Optimizing freezing time for low-moisture part-skim mozzarella cheese using multivariate and genetic algorithm approaches

This study applied multiple linear and nonlinear regression models to predict optimal freezing time for low-moisture part-skim (LMPS) mozzarella cheese and integrated these models with a multiobjective genetic algorithm (MOGA) and the TOPSIS multi-criteria ranking method. Cheese slabs were frozen at –18 °C across a wide range of time-to-freeze (TTF) values, and textural and rheological properties were evaluated using Texture Profile Analysis and temperature-dependent viscoelastic testing. Freezing significantly altered hardness, springiness, and chewiness (p < 0.0001), with the greatest reductions in chewiness occurring at short TTFs. Although a 19-minute TTF produced high elasticity, this condition reduced meltability and was therefore not suitable as an optimal freezing point. The integrated regression–MOGA–TOPSIS framework identified a TTF range of 43–48 minutes as optimal, achieving balanced firmness for shredding, elasticity for stretching, and uniform meltability. Multivariate analyses, including PCA, hierarchical clustering, and Pearson correlation, revealed strong interdependencies among key quality attributes such as hardness, elasticity, chewiness, and meltability. Among the regression models, quadratic regression achieved the highest predictive performance (R 2 > 0.75) for most parameters. however, these values indicate moderate predictive capability, and the models should be interpreted within the experimental range when used for optimization. This study demonstrates an integrated predictive optimization framework for improving freezing protocols in LMPS mozzarella cheese processing.