Food and nutrition security assessment requires an adaptive analytical approach due to the multidimensional and temporal complexity of food systems. This study proposes a hybrid decision support system integrating Multi-Criteria Decision Making (MCDM) methods, namely Analytic Hierarchy Process (AHP) and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), with machine learning to evaluate and predict food security indicators dynamically. Panel data from West Java and East Nusa Tenggara for the period 2018–2024 were analyzed to capture structural and temporal characteristics. AHP was used to determine expert-based indicator weights, which were applied in TOPSIS to generate regional food security scores. These scores were subsequently modeled using machine learning with temporal feature engineering, including lag variables and rolling statistics, and evaluated using time-series cross-validation. The results reveal a strong negative correlation (−0.7398) between AHP weights and machine learning feature importance, indicating complementary expert-based and data-driven perspectives. Ridge Regression achieved the best predictive performance with an R² of 0.9983 on training data and 0.8186 under cross-validation. East Nusa Tenggara outperformed West Java in TOPSIS scores (0.4829 vs. 0.4626), highlighting the importance of food stability and utilization. This study advances Informatics by enabling dynamic and adaptive food security decision support.