Incorporating CUSUM Procedure with Machine Learning to Detect Proxy Testing

• In Event: Electronic Board Session IV

Abstract

This study aims to detect proxy testers in repeated testing using machine learning techniques incorporating a statistical process control procedure, cumulative sum (CUSUM) with weighted residuals between observed and expected item scores. Results show that SVM and NN with CUSUM incorporated have a higher sensitivity and lower false positive rate.

Authors

• Guangyun Liu

• Xuechun Zhou, ASCEND LEARNING