Bee Swarm Optimization in Model Specification Search: A Simulation Study

This study extends the Bee Swarm Optimization (BSO) algorithm for model specification search in psychological test construction by adapting it from bifactor to correlated factor models. Using simulation studies under varying structural complexities and noise conditions, the performance of BSO was evaluated using multiple optimization criteria, including global and local model fit, and item retention. The results demonstrate that BSO consistently identifies an optimal model across all conditions, even with a small number of seeds. This data-driven method offers a promising alternative to traditional item selection methods, contributing to the refinement of measurement model development.