Machine Learning Methods for Predicting the Performance of Online Courses

The report discusses the development of a model for predicting the effectiveness of online courses based on activity in a virtual learning environment. The study is based on data from seven courses provided by The Open University of the United Kingdom. The analysis of the results shows that the best methods for predicting course success are random forest and gradient boosting. The prediction accuracy reaches 76.2%, with an f1-score of 60.2%. All methods used demonstrated good results, confirming the possibility of prediction based on available data. Thus, the presented study offers a new approach to assessing student performance in online courses using machine learning methods. The authors were able to increase the accuracy of predicting the success of online course completion by combining different prediction methods and adjusting their parameters.