Similar Users or Similar Items? Comparing Similarity-Based Approaches for Recommender Systems in Online Judges

• Published in: Case-Based Reasoning Research and Development pp. 92 - 107
• Main Authors: Caro-Martinez, Marta, Jimenez-Diaz, Guillermo
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Electoral System; Online Judge; Problem-based Approach; Similarity Metrics; Unweighted Metrics
• ISBN: 3319610295; 9783319610290
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-61030-6_7

Online judges store hundreds of programming problems but they lack recommendation tools to help users to find relevant problems to solve. In this paper, we extend the exploration of the use of the implicit knowledge derived from the relationships created between users and problems when the users submit their solutions to the online judge. Inspired by collaborative filtering techniques, in this work we compare a user-based and a problem-based approach, both supported by node similarity metrics coming from social network analysis, and we study the inclusion of voting systems in order to rank the problems that best fit for a user in the online judge. Our experiments reveal that the selection of the highest-performing similarity metric is determined by the recommendation method. We also show that the user-based approach outperforms the problem-based approach only when the proposed voting systems are used.