A curriculum recommendation system using a vector database for challenge-based learning

This paper proposes a methodology and describes the implementation of a graph-based curriculum recommender system for flexible, personalized, challenge-based learning. The central element of the proposed recommendation process is an algorithm that uses mixed integer programming to optimize curriculums based on multiple constraints, including logical sequencing and credit accumulation requirements. A procedure for assessing grades based on preliminary clustering of students using the k -nearest neighbors method is proposed. The key to implementing the proposed approach is generating performance vectors and estimating grades for tasks that have not yet been completed by students, which is necessary for personalizing learning recommendations. The architecture of a recommender system based on a vector database, the features of the algorithm for generating recommendations, and the results of assessing the degree of accuracy of estimates are described; generated synthetic data were used to conduct computational experiments.