A Personalized Recommendation System for English Teaching Resources Based on Learning Behavior Detection

Abstract

When confronted with a plethora of resources, many students struggle to quickly filter out the content that is relevant to them. Because there are many English teaching resources and it is difficult to accurately recommend suitable teaching resources for students. Therefore, in this paper we suggest a personalized recommendation system for English teaching resources, which is founded on learning behavior detection. To begin with, a spatiotemporal convolutional network is introduced to effectively identify students’ online classroom behavior, and a global attention module is added to increase the model’s ability to learn global feature information. Furthermore, the identified characteristics of student behavior are incorporated into the recommendation module. Similarly, the differential evolution (DE) algorithm is implemented to the smoothing factor and kernel function center of a generalized regression neural network (CRNN) for resource recommendation mode, while taking into account the strong dependence of the GRNN training effect on the smoothing factor and the kernel function center. The smoothing factor and offset factor are optimized and solved, and the optimized smoothing factor and offset factor are used to recommend GRNN resources. Experiments show that the approach described in this work first has a high precision (i.e., 90.98%) in behavior recognition, and second, the recommendation performance is superior to both of the comparison algorithms (i.e., 85.23% and 78.33%), resulting in better resource recommendation accuracy. The fundamental goal of this work is to deliver several important guidelines for the informatization and intelligence of the English educational resources and services.