<![CDATA[Early detection of toddler nutritional status is an important step in preventing the long-term effects of stunting. However, manual assessments are often inefficient and inconsistent, requiring a more accurate computational approach. This study aims to improve the performance of the Support Vector Machine (SVM) model in classifying the nutritional status of toddlers through hyperparameter optimization using Particle Swarm Optimization (PSO) and Genetic Algorithm (GA). A dataset of 120,999 entries was obtained from the Kaggle platform and processed through data cleaning, normalization, and categorical encoding stages. A baseline SVM model was built as an initial benchmark before optimization. The results of the experiment showed that the PSO method produced the best performance with accuracy, precision, recall, and f1-score of 0.995, surpassing SVM without optimization (0.950) and SVM optimized by GA (0.990). These findings indicate that PSO is capable of exploring the solution space more effectively based on experimental results showing more stable and gradual convergence on the learning curve, with a smaller gap between training and validation values, and a more structured solution search process compared to GA, thereby improving model stability and accuracy. This research contributes to the development of a more accurate classification model for toddler nutritional status that can be integrated into decision support systems in the health sector, which can also be applied to improve efficiency and accuracy in data-based decision making, particularly in the context of assessing toddler nutritional status. This contribution has a significant impact on advancing the application of metaheuristic methods in various informatics applications.]]>
