[4N2-GS-10] AI application

[4N2-GS-10-04] Automated Test Assembly using Maximum Weight Clique Aalgorithm

〇Mizuho Kadowaki¹, Kazuma Fuchimoto¹, Maomi Ueno¹ (1. The University of Electro-Communications)

Keywords:automated test assembly, parallel test, maximum weight clique problem, item response theory

E-Testing of the global standard for computer-based testing is necessary for the automated test assembly of parallel test forms, by which each form has equivalent measurement accuracy but with a different set of items. An important task for automated test assembly is to assemble as many tests as possible. To resolve this problem, Fuchimoto et al. (2022) proposed automated test assembly using the maximum clique problem with integer programming, which is known to assemble the greatest number of uniform tests with the highest measurement accuracy. This method resolves the tradeoff between the number of assembled tests and the equality by manual hyperparameters. In this paper, we propose a maximum weight clique method to determine hyperparameters automatically. Specifically, the proposed method constructs a graph in which the vertices, the edges, and the weights of vertices represent tests satisfying the test constraint, the satisfaction of overlapping, and the difference between the target measurement accuracies and the measurement accuracies of tests. Then, the proposed method seeks a maximum weight clique in the graph. Numerical experiments demonstrate that the proposed method reduces the difference in measurement accuracies between tests without decreasing the number of tests.