Arabic Predicting Course Difficulty in Online Education Using Machine Learning

Abstract

The booming proliferation of online educational platforms over the last several years has transformed the landscape of learning resources. On one hand, this transformation has generated many promising opportunities to optimize course delivery and student performance, while, on the other hand, it has posed various dilemmas, one of which is accurately predicting the course level of difficulty. Since the assessment of course difficulty using traditional methods varies greatly and is often determined subjectively, it is difficult for learners to assess whether they have the necessary skills and knowledge to manage the course or not. To resolve this problem, in the present study, we develop a machine learning-based framework to predict course difficulty levels on Coursera Course Dataset. Our contributions include the employment of three strong classifiers, which we compare to one another: GB, RF, and XGBoost. We also conducted a considerable amount of preprocessing, such as missing values, categorical variables encoding, and SMOTE for balancing the dataset. The evaluation results demonstrate the superiority of the XGBoost model with an accuracy of 96.4% and excellent precision, recall, and F1 scores for all classes. The implications of this study include not only its potential for enhancing course recommendation systems and personalizing online education but also for further refinement by introducing more features and real-time predictions.