Bayesian Adaptive Diagnoser of Bug Rules with Tree Structure Model

T. Ohmori and K. Shigemasu (Japan)


Bug Rule, Bayesian Approach, Adaptive Testing, EVSI


In this paper, we propose a new diagnostic system to identify the bug rules used by a student while solving arithmetic problems; this system is based on the Bayesian approach. In diagnosing tests, it is advantageous if the testing system requires a lesser number of test items for identifying the students’ abilities. During the diagnosis, this system automatically selects the optimal set of test items. The system assumes that bug rules are represented in the form of a tree-structure model; it calculates the posterior probability of each student using each of the possible bug rules. The validity of the system was verified by carrying out simulations using real test data.

Important Links:

Go Back