Course Video Recommendation Algorithm Based On Differences In User's Long And Short Interest In Educational Scenes

In today's era of information explosion,personalized recommendation services have become one of the main channels for people to obtain valuable information for themselves.Major companies are studying how to provide users with more accurate and personalized recommendations.In recent years,recommendation algorithms based on user history sequence information have become an important research direction in the recommended field.Sequence recommendation based on cyclic neural network and convolutional neural network is currently used more sequence recommendation models.However,the current sequence recommendation algorithm still has some shortcomings: the current sequence recommendation algorithm lacks the long-term preference learning for the user,and the recommendation result of the user preference change algorithm will generate redundancy.This paper will focus on some applications of sequence recommendation in educational scenarios,and solve some problems existing in the current scenario to meet user needs.The main work of this paper is as follows:Design a graph-based walk-through representation learning algorithm based on collaborative filtering.The widely used graph walk alogrithm is very stable and robust to the learning behavior of user behavior under large-scale data sets.In this paper,the collaborative filtering algorithm is introduced to preprocess the user behavior and filter the abnormalities and mutations in the user behavior.Further improving its accuracy,the experimental results show that the graph-based walk based on collaborative filtering indicates that the learning algorithm can better learn the vector representation of user behavior.Design the user length and interest difference model deep learning algorithm: The sequence recommendation deep learning alogrithm based on user behavior mainly uses the short-term behavior of the user to learn and represent,lacks the representation of the longterm behavior of the user,and has no concern about how the long-and short-term behaviors are integrated.This paper introduces the Attention-based model to represent the long-term and short-term behavior of users,and uses the long-term and short-term behavior differences to conduct selective learning methods,which solves the need to identify user interest changes in educational scenarios.Design and implementation a lightweight recommendation system: The recommendation system includes algorithms model,front and back,design of the database and some processing methods.At present,the system has been operating stably in the actual production environment to provide services for related products.