Predicting Students Performance Based on Their Reading Behaviors

E-learning systems can support students in the on-line classroom environment by providing different learning materials. However, recent studies find that students may misuse such systems with a variety of strategies. One particular misused strategy, gaming the system, has repeatedly been found to negatively affect the students’ learning results. Unfortunately, methods to quantitatively capture such behavior are poorly developed, making it difficult to predict students learning outcomes. In this work, we tackle this problem based on a study of the 567,193 records of the 71 students’ reading behaviors from two classes in the academic year 2016. We first quantify the extent to which students misused the system and then predict their class performance based on the quantified results. Our results demonstrated that such misbehavior in the E-learning system can be quantified as a probability and then further used as a significant factor to predict students class learning outcomes with high accuracy