Features Correlation Analysis and Classification based on EdNet User Online Learning Behavior

Abstract

Online learning has become an increasingly popular way of learning. In order to improve users' online learning experience and learning effectiveness, analysis on users' online learning behavior has become a hot issue in the field of education big data. Based on the EdNet data set, this research randomly selects some users, counts these user's answer scores, elapsed time and other learning behavior data, and then extracts feature. Meanwhile, features of questions difficulty are calculated on the basis of the EdNet raw data, and construct user completion difficulty features. By extracting and constructing the features of users' learning behavior, a random forest model is used to classify and predict the user's level. The experimental result shows that, on the issue of classifying user's level, the user completion difficulty features are conducive to model performance. It also confirms that the features of questions difficulty has a great relationship with user's learning effectiveness. Moreover, this research gives some suggestions for users to improve their learning performance.