Personalized Recommendation Service of Educational Media Resources Based on Multi-dimensional Feature Fusion

Abstract

Video is an important carrier of all kinds of information on the Internet. More and more students choose to watch online learning videos from online learning platforms with various forms of intelligent terminals. To allow students to get the videos which they really need and are interested in from the mass videos, it is necessary to improve the existing personalized recommendation algorithm. Content-based recommendation method, collaborative filtering-based recommendation method and hybrid recommendation method often have the problems of loss of feature information or cocoon. To this end, this paper studies the personalized recommendation service of media educational resources based on multi-dimensional user features. This paper constructs a multi-feature candidate set based on student-video interaction, extracts and splices category feature, label feature and semantic feature of Internet media educational videos, and generates the representation of students’ learning preference feature. Based on the improved forward and backward recurrent neural network, this study constructs a recommendation model of Internet media educational videos oriented to students' preference points. The experimental result verifies the validity of the model.