A Hybrid Model For Predicting E-learning Course Dropout Rate For Post Graduate Students

Abstract

In universities all around Kenya, e-learning has grown in popularity, especially for postgraduate programs. Students now find it simpler to access education from any location at any time thanks to the use of technology in the delivery of courses and academic resources. However, dropout rates continue to be a serious issue despite the many advantages of online learning. For a variety of reasons, including a lack of desire, insufficient assistance, and trouble understanding the course materials, students withdraw from online courses. Dropouts drive up educational institutions' average cost per student because it typically costs more to retain a possible dropout than to enroll a new student. The rise of online learning is hampered by the prevalence of school dropouts, which waste the student's initial time and financial investment. Low graduation rates that follow high dropout rates will surely damage the standing of educational institutions in the community and eventually result in a downward loop of declining government support. To lower the dropout rate, online educational institutions can employ this technology to quickly spot probable dropouts and put retention measures in place before the dropout behavior takes place. The study's objective was to create a hybrid machine learning prediction model for postgraduate E-learning students who drop out utilizing the Support Vector Machine and Random Forest algorithms to improve prediction accuracy. The researcher employed a descriptive survey and an experimental study approach. The research methodology will be appropriate because the researcher trained the Dropout Prediction Detection model using a machine learning technique. In 2024, 61.7% of students are expected to graduate. With the aid of the data, the researcher was better able to determine whether students had spent more time studying than was anticipated. 62.5% of the respondent's price posed the most challenge to finishing the investigation, however 37.5% of the fee posed no issue. In order to prepare students for postgraduate study, 68.3% strongly agreed that undergraduates should be taught research techniques, and 29.2% also agreed. A 100% accuracy rate for forecasting student dropout was demonstrated by the hybrid model. By using machine learning to predict student attrition, educational institutions have a ground-breaking chance to effectively address this pervasive problem. The study also recommended that the students choose a study strategy that would best fit their schedules in order to prevent unneeded stress from juggling numerous tasks at once. Deep learning models can be strengthened by techniques like Synthetic Minority Over-sampling Technique to handle the unbalanced datasets typical in dropout prediction problems