Leveraging neural network for online learning performance prediction and learning suggestion

Abstract

Learning performance analysis is such a research field that draws much attention from researchers though it has just been emerged in recent years. On the one hand, analyzing learning behaviors can help learners to choose their learning methods and allocate their study time in a more appropriate way. On the other hand, learning analysis can provide valuable feedbacks for teachers and administrators to improve teaching efficiency and quality. This paper studies and analyzes more than 640,000 learning data from the MOOC platform edX. A tree-based model along with an information gain measure is applied to identify the usefulness of data features. A back-propagation neural network model is further adopted to train data and achieve a prediction model of learning performance. In addition, a genetic algorithm calculates learning score conditions and return feedbacks as suggestions to learners. Experiment results demonstrate the effectiveness of the utilization of the methods in the predication of online learning performance.