A fast convergence Computerized Adaptive Testing system based on multi-agent system

A multi-agent median computerized adaptive testing (CAT) system is put forward, which is based on the three-parameter logistic model (3PLM) of item response theory (IRT) and a multi-agent item pool. The multi-agent item pool is proposed to help the non-professionals system choose valuable items interactively and conveniently . Aiming at the problems in the ability estimate of multi-agent adaptive testing, rapid convergence can be realized greatly if logistic median algorithm is applied to calculate maximum likelihood estimation (MLE) instead of the most commonly used Newton_Raphson iterative algorithm. Simulated experiments are conducted with the research object of multi-agent three parameter logistic median model. The result shows that this CAT system has more superior function and lower calculation quantity, with fast convergence and the accuracy of ability estimate, which will be adapted further in modern test system.