Machine Learning Algorithms for Early Predicting Dropout Student Online Learning

Online learning is different from offline learning in the classroom with supervision from the lecturer. Online learning using the Learning Management System (LMS) media requires high awareness from students because their learning activities are not supervised, they are free to study wherever and whenever, so they need to manage and control their own study time without the help of lecturers or administrators. This is one of the causes of the high dropout rate among online learning students, so it is very important for lecturers and administrators to support students in a timely manner to avoid the risk of dropping out. This study uses access log data recorded in the LMS and student statistical information and calculated data and aims to present a suitable predictive algorithm for dropout early prediction systems for online learning students using machine learning. Of the 4 algorithms used, the highest recall value is in Naive Bayes (1), the highest precision is in Logistic Regression with Lasso (1), while the highest accuracy value (0.99) and F1score (0.97) are obtained from the Support Vector Machine which has value equal to Logistic Regression with Lasso. In general, the early dropout prediction model will allow lecturers and administrators to focus on students who have the potential to dropout and take quick action to improve their learning performance so as to reduce the number of student dropouts.