Deep Learning-Based Early Dropout Prediction in University Online Learing

With the global transition of universities to online education due to COVID-19, the high dropout rate in online learning has become a critical challenge for higher education institutions. To address this issue, this study aims to develop a deep learning model for early dropout prediction in university online education. The proposed model was built by collecting and analyzing daily learning history data stored in the Learning Management System (LMS). Unlike previous studies that primarily relied on data collected at the end of the online learning period, this study analyzes students' behavioral data over time to more accurately identify students at risk. The research utilized data from a cyber university located in Seoul, South Korea, including approximately 30,574 student records and 12,014,610 learning history entries from the academic management system. To validate the model’s performance, data from the following academic year, which was not used for model training, was employed. The study compared the effectiveness of traditional machine learning methods with deep learning techniques (DNN and LSTM). Specifically, it proposed the LSTM-DNN model, which effectively learns both static learner information data and sequential learning history data. The results demonstrated that the LSTM-DNN model achieved a prediction accuracy of over 92%, confirming its effectiveness in providing real-time dropout risk assessments and predictive insights. Ultimately, this study proposes a novel approach to integrating real-time dropout prediction services into university Learning Management Systems (LMS), thereby contributing to student retention and academic success in online learning environments.