An approach to the personalized learning objects recommendation problem as a set covering problem using ontologies and metaheuristics

Abstract

Recommender Systems are extensively utilized in e-commerce platforms, such as sales websites and Netflix, to intelligently suggest products, movies, and series tailored to the user’s preferences. In the context of education, the key challenge for these systems is to provide personalized recommendations of educational content that align with students’ needs, considering their knowledge levels, learning styles, and cognitive preferences. This work implements a recommender system designed to suggest learning objects across various areas of knowledge, integrating small learning objects, called interventions, such as definitions, examples, and hints. To personalize these recommendations, the Learning Objects Recommendation Problem is formulated as a set-covering problem, which belongs to the class of NP-Hard problems. A heuristic search-based algorithm was proposed and compared with other metaheuristics, resulting in a promising approach to solving this problem, as demonstrated by the results. The proposed solution aims to minimize the challenges of cold-start and rating sparsity, common in traditional recommender systems, by using advanced collaborative filtering techniques and an ontology that models the students’ needs, knowledge, learning styles, and search parameters. Additionally, the recommender system was implemented with a chatbot and tested for recommending content on the C programming language for first-year students of the Computer Science course, using gamification to alleviate possible pedagogical difficulties in the teaching-learning process.