| Searching for courses on a topic in a university database or listing of courses can prove difficult. Strictly in Montreal universities, the number of courses range in the thousands. The problem is exacerbated by the fact that many courses are multidisciplinary. For graduate students in particular, who should look for courses on a topic related to their research, it implies that defining their course plan can be a difficult process that requires some assistance. Even when a course that is relevant is found, it often is not offered in the right semester or it is filled to capacity. Therefore, a system that provides a means of finding courses based on their similarity would prove very useful.;A number of systems have been developed to provide course recommendations to students, but we aim to define an approach that is solely content-based, using the similarity of course descriptions. The algorithm is based on the vector-space model of the term-document matrix.;This thesis presents the FCRC approach (content-based course recommender) which offers recommendations based on course similarity measures.;Results show that the similarity measured on the Dice coefficient between document vectors offers relatively complete and precise recommendations. The cosine is also a good measure of similarity. In general, the first 5 of 10 recommendations are relevant based on this approach, and the recall rate is close to 100%. |
