Context-Aware Recommendation-Based Learning Analytics Using Tensor and Coupled Matrix Factorization

Abstract

In higher education, student retention and timely graduation are enduring challenges. Educational, advising, and counseling innovations and interventions are needed to address these challenges. With the rapidly expanding collection and availability of learning data and related analytics, student performance can be accurately monitored, and possibly predicted ahead of time, thus enabling early warning and degree planning `expert systems' to provide disciplined decision support to counselors, advisors, educators -- and even help guide students in semester-to-semester course selection. Previous work in educational data mining has explored matrix factorization techniques for grade prediction, albeit without taking contextual information into account. Temporal information should be informative as it distinguishes between the different class offerings and indirectly captures student experience as well. To exploit temporal and/or other kinds of context, we develop three approaches that leverage side information besides historical grades under the framework of Collaborative Filtering (CF). Two of the proposed methods build upon Coupled Matrix Factorization (CMF) with a shared latent matrix factor. The third method utilizes tensor factorization to model grades and their context. For each method, the latent factors obtained using matrix/tensor factorization lead to a compact model which we use not only to predict the unseen grades, but also the associated contextual information. We evaluate these approaches on grade datasets obtained from the University of Minnesota. Experimental results show that quite accurate prediction is possible using even simple models, while more advanced approaches outperform the prior art in predicting randomly missing entries.