Context aware practice problem recommender system in online judge using machine learning techniques

Abstract

Recommender systems in electronic portals reduce user effort and improve user engagement. Various portals, such as e-commerce, e-learning, video streaming, music, advertising, etc., are using recommender systems to help users to choose pertinent items based on their current context. The current context of a user refers to various factors, and those factors are not common across all the portals. For example, the current context signifies user needs and preferences in e-commerce, user mind-sets in video streaming portals, and topics of interest and knowledge level in an e-learning portal. Many research works have focused on context-aware e-learning recommender systems based on the user s knowledge level, learning path, topics of interest, and performance. Online Judge (OJ) is an e-learning platform specifically designed to learn programming by practice. Since OJs have a large number of problems, it is hard to choose the right problem to solve. If the chosen problem is very hard to solve, the user might lose interest in solving problems. To recommend practice problems in OJ, the recommender system should consider the user s current context (knowledge level, problem-solving ability, topics of interest, learning path, and performance). The traditional Collaborative Filtering (CF), Content Based Filtering (CBF) and other recommender systems designed for e-learning are inadequate for this recommendation. Another challenge in OJ recommender system is availability of user ratings. Generally, most of the users have not shown interest in giving ratings to problems, as they do to items in e-commerce. User rating information is not sufficient to implement traditional recommender systems newline