Providing personalized learning guidance in MOOCs by multi-source data analysis

Although millions of students have access to varieties of learning materials in Massive Open Online Courses (MOOCs), many of them feel lost or isolated in their learning experience. One of the potential reasons is the lack of interactions and guidance for individuals. Since MOOC students have diverse learning objectives, we propose to design different strategies for those students with different engagement styles. In this paper, we provide personalized learning guidance for MOOC students based on multi-source data analysis. We first conduct content analysis to identify key concepts in the courses. We then propose two structured model to evaluate student knowledge states by their quiz submissions. We also study on student learning behaviors and design a dropout prediction system. The experiments show the effectiveness of our algorithms and we discuss on the result both quantitatively and qualitatively. Last but not least, we employ a Web application of online student assessment service for both students and instructors, in order to display student learning states and provide suggestion for individuals.