Reengineering the Feature Distillation Process: A case study in detection of Gaming the System

As education technology matures, researches debate whether data mining (EDM) or knowledge engineering (KE) paradigms are best for modeling complex learning constructs. A hybrid paradigm may capture strengths from both approaches. In particular, recent work has argued that successful data mining depends on thoughtful feature engineering. In this paper, we explore the use of cognitive modeling (a form of knowledge engineering) to enhance the feature engineering process for detectors of gaming the system, one of the most studied complex constructs in EDM. Using this construct enables us to measure the extent to which our techniques improve performance over previous models.