Abstract:

It is already well-known that technological developments have an undeniable effect on education. At most points, technological developments have a direct share in the improvement of the educational process. Early warning systems are one of these technological touches that affect education positively. Early warning systems, as an enhancement of learning analytics systems, are used to better support students based on their behavior and performance and identifying potential at-risk situations by collecting student data through technologies such as learning management systems or databases, which already have students previous signs of progress. Data, such as student participation, behavior and course performance constitute the basic input of an early warning system. Additionally, an early warning system does not require any special effort by teachers or any other participants rather than the existing data. It analyzes the risk status and achievement status of the participants for their future performance and presents them as a warning.

This study aims to identify students at-risk by using the simple Gradual At-risk (GAR) predicting model in the Computer Structure course in the Universitat Oberta de Catalunya (UOC) and to provide early feedback based on the chance to pass the course. Computer Structure course with 249 enrolled students expands the knowledge of the hardware components in the undergraduate bachelor of Computer Science. The course has four assessment activities (AA) during the semester timeline, and the early warning system is capable to identify potential at-risk students from the very first activity with an accuracy of the 73.49%. This study extends a previous one, which was aimed to develop an early feedback prediction system for learners based on data available in our institutional datamart (known as the UOC Datamart). The results of this study will demonstrate the effectiveness of the early warning system to identify at-risk students based on the GAR model and by using the Green-Amber-Red risk classification.

Keywords:

Predictive models, at-risk student, early warning systems, online learning.