All Times EDT
Keywords: Logistic Regression, Real-time Analysis, Dig Data Analytics, Educational Analytics
Early prediction of academic at-risk students is important because it can maximize the opportunity for effective intervention to increase the students’ retention rate. In this study, we examined students’ performances through key courses, predicted students’ academic risk before the semester begin using students’ demographic and academic background information from Student Information System (SIS) data and then validated the consistence of our initial prediction throughout the semester using additional data from Learning Management System(LMS). First, we built a base-line model by considering only SIS data without students’ course performance data before the semester begin. In this base-line model, we identified an initial set of academic at-risk students. Then, a progressive model was built weekly with additional student learning performance data collected through LMS system to validate the accuracy of base-line model. Training dataset was based on student data of key courses in Fall 2017, and test dataset was the same courses’ data in Fall 2018. Model evaluation showed the baseline model had predict at-risk students with a high level of precision and the weekly progression models had confirm the consistency of the baseline model. In addition, our weekly progressive models provided additional insight on actionable factors leading to the development of better prevention strategies to reduce students’ academic risk. Our findings can be beneficial to university administrators on helping academic at-risk students to improve their course performances, to decrease their graduation time, and to prevent them from drop out.