Module recommendation systems that accurately capture students’ preference and recommend modules accordingly could save millions of university students from the stress over module selection. Yet, limited studies focus on the application of recommendation systems (RS) in the niche area of module recommendation.
In this project, we explore ways to fill in this research gap. So far, we have worked with module enrollment history from three faculties of the National University of Singapore (NUS), including School of Computing (SOC), Faculty of Arts and Social Sciences (FASS) and School of Business. Grade-enhanced Multi-layer Perceptron Model (GEMLP) is proposed to learn both general interest and grade consideration and finds the optimal balance between these two factors by incorporating two weightage embedding layers. Grade-to-interest Ratio is computed based on values for these embedding layers which quantitively reflects how students weigh interest and grade consideration during module selection. Results show that students across faculties indeed behave differently when selecting modules.
For next step, we are in the process of applying for more faculty data and module data to augment the current dataset. Current analysis will then be applied at a finer-grained program level. Other directions will be explored to expand the project as well.