Since cheating in testing is not only an ethical issue but also adversely affects the validity of tests, techniques for detecting cheating are a necessary part of test operations. In this study, to address this need, simulation studies were conducted to identify individuals who cheated on a computerized adaptive test by utilizing prior knowledge that was not their own original ability or knowledge. Machine learning techniques were employed using multiple person-fit indices as explanatory variables, with a focus on minimizing the number of false positives. The findings indicated that the detection of cheaters was more effective as the number of leaked items increased. Additionally, the results were decomposed to identify the contribution of each explanatory variable, revealing the effectiveness of using multiple person-fit indices as explanatory variables in detecting cheaters.
