Research on Learning Resource Recommendation Algorithm Based on User Demand Evolution

Abstract

With the continuous development of online education platforms, a vast number of learning resources have brought challenges to users’ choices. It is feasible to introduce the recommendation algorithms in e-commerce directly into the online learning platforms to recommend courses for learners. However, they often ignore the dynamic evolution of learning needs, resulting in partially ineffective recommendations. To solve this problem, we propose a method called LR-DE, which takes the change rule of user needs into account. It first exploits gated recurrent units to extract the users’ learning interest states at each time step. Then, to address the phenomenon of interest incoherence in behavior sequences, we introduce the attention mechanism to upgrade the update gate in GRUs structure and construct a new network structure called AUGRU. At the same time, the bilinear feature interaction is used to calculate the correlation scores between the click histories and the candidates so as to capture the co-occurrence relation between the two courses. Experimental results show that our method is superior to the existing methods in learning resource recommendation tasks and can effectively improve recommendation accuracy.