So Happy Together? Combining Rasch and Item Response Theory Model Estimates with Support Vector Machines to Detect Test Fraud

An Ensemble-Unsupervised-Learning-Based Approach for the Simultaneous Detection of Preknowledge in Examinees and Items when Both are Unknown

“…In terms of preknowledge detection, supervised learning involves building a model that predicts the compromise status of items/individuals within a dataset in which compromise status is known, and doing so in a manner that is most likely to generalize to future datasets in which compromise status is not known. Using a variety of item features under Item Response Theory (IRT) models as input variables, Thomas (2016) used support vector machines to identify CI on a certification exam for which approximately 60% of the items were suspected of being compromised. Zopluoglu (2019) used examinees flagged for preknowledge to train an Extreme Gradient Boosting algorithm to detect EWP in large-scale testing.…”