Using Data Mining to Predict Student Disengagement in Tutoring Systems

• In Event: 47.033 - The Road to Validity Through Student Motivation, (Dis)Engagement, and Aberrant Behavior Detectors

Abstract

This research seeks to identify the accuracy with which student disengagement can be predicted using the data generated within a tutoring system, demonstrating how easily process data can be used for prediction and to explicate student behavior. A variety of variables were used and three predictive models were explored. Neural networks resulted in the highest predictive power, with 63% of student drop outs accurately predicted.

Author

• Elizabeth Adele Patton, Curriculum Associates