Study on the Influencing Factors of Online Learning Effect Based on Decision Tree and Recursive Feature Elimination

With the popularity of online learning, more and more researchers have attached great importance to the relationship between learning effect and influence factors in the online courses. In literature works, Logistic Stepwise Regression algorithm is the most used method. But this method has limitation in run time especially when the dimension of data is large. Besides that, it can't rank the importance of factors. Aiming at the above shortcomings, this paper proposes a novel approach to analyze the influencing factors of online learning, which is based on the combination of decision tree and recursive feature elimination. Firstly, the feature sorting algorithm is based on decision tree to conduct the preliminary screening, which is to form a candidate feature set. Then, recursive feature elimination is used to rank the candidate features by their importance. At this stage, Logistic Regression (LR), Support Vector Machine (SVM) and Decision Tree (DT) models are used separately to obtain each collating sequence of importance. By averaging these collating sequences, the final importance ranking of candidate features is achieved. Finally, an experiment is carried out on the Open University Learning Analytics dataset, and the results show that learning behavior has an important impact on the learning effect. Positive learning behaviors can lead to better learning effect.