Evaluation for the Test Quality of Dynamic Question Generation by Particle Swarm Optimization for Adaptive Testing

Abstract

This study discusses the dynamic item-selecting mechanism used in the testing method of Adaptive Testing by Particle Swarm Optimization (PSO). By setting weights of knowledge, we have designed adaptive questions and have evaluated an item’s difficulty and discrimination level based on the Item Response Theory (IRT) to see if testees’ correct responses match their ability and reach the purpose of adaptive testing. The study designs three experiments to evaluate the quality of the adaptive testing. Namely, they are item discrimination and testee’s ability analysis, item exposure rate control and respective test quality.