Comparative Analysis of the Quality of Predicative Models for the Online Educational Platform

This research paper is mainly concerned with studying aspects of applying various techniques of Data Science in the educational context – this new direction in the field of data analysis has been called the Educational Data Mining. The purpose of the research work is to provide a comprehensive review and selection of the most productive and precise classification models for forecasting student academic performance on the online educational platform. The main tasks of a research conducted are as follows: 1. to provide a review of the practices in relation to applying EDM methods in predictive modeling tasks; 2. to develop versatile classification models to predict student’s academic performance; 3. to conduct a comparative analysis of multiple constructed classifiers. In predictive modeling, anonymized data from courses presented at the Open University were used. It contains demographic, performance and learning behavior data about more than 32 thousand students, learning for 7 online courses. Even though numerous studies focused on applying classification for predicting academic performance are already conducted, some crucial modeling aspects were neglected. List of this aspects includes overfitting detection, hyperparameter optimization, choosing suitable model evaluation metrics, combining classifiers and various machine learning methods. Taking into account outlined vulnerabilities, several predictive models were constructed, ranging from widely used and rather popular classifiers in EDM studies to the most advanced and sophisticated techniques, the application of which was not previously considered in the educational context. Based on outcomes of the study, the most effective and promising models were identified. Consequently, the result of this research paper might be valuable in the field of online education, where advanced predictive models are potentially able to be a part of decision support systems, suitable to improve global educational practice.