Personalised course recommender

In general, course-recommendation systems retrieve results based on a learner’s query. Hence, they are limited to making recommendations based on the keywords of that single input and do not take into account additional information regarding the user. 

This research investigated the possibility of providing a proof of concept by showcasing the benefits to the field of e-learning that a personalised course-recommendation system based on user profiling could offer. This was attempted by using artificial intelligence (AI) and machine learning (ML) techniques. The proposed system would be able to generate a number of recommended courses based on their personalised profile, without having to search for a specific query. Additionally, the predicted success rate of completing each recommended course would be displayed.

The recommendation system was implemented using a hybrid approach, in order to minimise the main issues present in most online recommenders. Therefore, the system has been developed using a combination of content-based and collaborative filtering. The content-based filter classified items using keywords to recommend similar items, whereas the collaborative-based filter classified users into clusters of similar types and recommended courses based on the cluster in which the user was classified. By combining the list of recommendations of both approaches, the anticipated result was an accurate and proficient recommendation system that would be able to generate the top recommended courses for each user, based on their own unique profiles.
Finally, a success-rate predictor was implemented using the collective information of the user profile and the course details, to calculate the student’s probability of completing the course. This feature was intended to assist potential students in selecting courses based on their current academic level and to save their time and funds in the event of a course being predicted to have a low success rate. The prediction was calculated by going through the course details and matching course requirements with the user profile.

Figure 1. Outline of the recommendation system

Figure 2. Sample results generated by the recommendation system

Student: Britney Vella

Supervisor : Prof. Matthew Montebello