Abstract
Massive open online courses (MOOC) is a novel educational model emerging in recent years, which is assisted by advanced techniques such as cloud computing, big data and Cyber-Physical Systems (CPS). Through adequate analysis assisted by big data, the quality of education is expected to be extensively improved. Unfortunately, the MOOC data are not fully utilized for educational innovation, because the existing research focuses on the data generated in the online learning but neglects other related data, such as the forum data. In this article, we propose a big-data-driven approach named TOLA for online learning evolution to discover students' learning pattern and guide courses improvement. Specifically, topic feature is extracted from MOOC forum through Latent Dirichlet Allocation, which is incorporated with other hybrid features. Through experiments, TOLA exhibits good performance in terms of complexity, efficiency and accuracy, facilitating the improvement of the quality of online education. (C) 2016 Elsevier B.V. All rights reserved.