Towards effective course-based recommendations for public tenders

Abstract

In this paper, we propose a recommendation model to assist users find relevant courses for public tenders. The recommendations are computed based on the user study activity at Atepassar.com, a web-based learning environment for public tender candidates. Unlike traditional academic-oriented recommender systems, our approach takes into account crucial information for public tender candidates such as salary offered by public tenders and location where the exams take place. Technically, our recommendations rely on content-based techniques and a location reasoning method in order to provide users with most feasible courses. Results from a real-world dataset indicate reasonable improvement in recommendation quality over compared baseline models - we observed about 11 precision improvement and 12.7% of recall gain over the best model compared - demonstrating the potential of our approach in recommending personalised courses.