With the support of the "Internet plus" strategy,Internet plus education was highly regarded as an important branch,which promoted the integration of information technology and education.Among them,online education was an indispensable way of education.The proportion of online courses in daily education activities increased rapidly,and the demand of teachers and students for online course resources also increased rapidly.However,massive curriculum resource data brought the problem of information overload,which made users face many challenges to obtain personalized these resources.In this context,personalized recommendation technology emerged as the times require and became an effective solution.Existing collaborative filtering algorithms analyzed users’ potential behavior preferences through users’ historical behavior data.In the Online Course Resource Recommendation problem,resource content contained a lot of user content preference information,which was of great significance to understand users’ needs more accurately,this paper proposed the LGCM(light graph convolution network and content similarity calculation mode).Collaborative filtering recommendation were realized based on the light graph convolutional network model,this model integrates the CSCM(content similarity calculation model)to achieve hybrid recommendation.The model analyzed the user’s potential behavioral preferences based on user behavior characteristics,analyzes the user’s online course resource content preferences based on the user’s content demand characteristics,and finally combines the two kinds of preference analysis to achieve personalized resource recommendation.This study first made collaborative filtering recommendation based on the lightGCN model component,and mines users’ potential behavior preference for online course resources through analyzing users’ behavior characteristics.The user resource behavior interaction matrix was convoluted by lightweight graph and combined by layers.The characteristic representation of user and course resource was generated and the inner product was calculated.The predicted score of user to course resource was obtained and the recommendation list was generated.Secondly,the CSCM component was introduced to mine users’ resource content preferences.The text features of online course resources were extracted,and the word2vec model was trained by using the resource database to transform the text features into word vectors.The distance between vectors was calculated to represent user preferences and content similarity between resources.Finally,the final recommendation result was generated by model fusion calculation.The data set was obtained through data processing,and a series of experiments were designed to verify the effectiveness and recommendation performance of the model.It was proved that the LGCM model can make personalized recommendation based on users’ potential behavior preference and content interest preference,and the recommendation performance was improved compared with the LightGCN model. |