Merged Tree-CAT: A fast method for building precise computerized adaptive tests based on decision trees

Abstract

Over the last few years, there has been an increasing interest in the creation of Computerized Adaptive Tests (CATs) based on Decision Trees (DTs). Among the available methods, the Tree-CAT method has been able to demonstrate a mathematical equivalence between both techniques. However, this method has the inconvenience of requiring a high performance cluster while taking a few days to perform its computations. This article presents the Merged Tree-CAT method, which extends the Tree-CAT technique, to create CATs based on DTs in just a few seconds in a personal computer. In order to do so, the Merged Tree-CAT method controls the growth of the tree by merging those branches in which both the distribution and the estimation of the latent level are similar. The performed experiments show that the proposed method obtains estimations of the latent level which are comparable to the obtained by the state-of-the-art techniques, while drastically reducing the computational time.