Power of Attention in MOOC Dropout Prediction

Abstract

The dropout rate of massive open online courses (MOOC) has been significantly high, which makes its prediction an important problem. In this article, we try to transfer the knowledge gained in the field of Natural Language Processing into the field of MOOC dropout prediction, due to the high similarity between them. More specifically, we attempt to study and show the powerful use of attention and conditional random field, both of which have been very popular architectures when solving NLP problems. A novel neural network structure is designed as the combination of these techniques. Extensive experimental results demonstrate that the proposed approach is effective.