Data Consideration for At-Risk Students Early Alert

Abstract

In recent years, leveraging big data technology to education domain draws attention. The retention rate of student on campus may be significantly improved if an early alert system based on data analysis is setup and the intervention is appropriately deployed. However, to the best of our knowledge, the existing proposed early alert systems take the whole participants in a group or one single class as analysis target. In this paper, we compare different strategies and methods that deal with the whole participants and groups in different classes to identify the key attributes to improve the accuracy of the proposed early alert systems. Our results in this study indicate that teachers in different classes may make use of different functionality of LMS. Different data volumes are collected in different classes. A robust early alert predictive model needs to take the above into consideration.