| The huge library of learning resources in current online education platforms makes users overwhelmed and time-consuming in selecting high-quality and well-suited resources.As an intelligent tool to deal with information overload,learning resource recommendation methods can reasonably analyze users’ preferences and recommend suitable courses for them according to their preferences.However,some existing learning resource recommendation methods on the one hand ignore the user interaction time information and cannot dynamically extract user interests,and at the same time have the problems of sparse data and missing data labels to effectively extract data correlation.On the other hand,the current learning resource recommendation system uses centralized data storage,which is prone to user privacy leakage and thus reduces the usage rate of the learning resource recommendation system.Therefore,to address the above problems,this paper uses deep learning methods to perform intelligent recommendation and combines block-chain storage technology to enhance the security of data in the recommendation system.The main research contents of this paper are as follows:(1)To address the problem that existing learning resource recommendation methods only consider explicit interactions and ignore the information implicit in the interaction sequence as well as fail to consider the interaction time.A time-aware lightweight graph convolutional network-based learning resource recommendation method is proposed.Firstly,the interaction data of users and courses are constructed in the form of bipartite graphs,and time interval information and relative location information are introduced in the embedding layer;secondly,a time-aware aggregator is designed to aggregate the user and course neighborhoods to realize the modeling of the implicit relationships;and then a multi-layer graph convolution is superimposed to extract the higher-order synergistic information of users and courses,which improves the recommendation accuracy.(2)To address the problems of sparse data and missing data labels in the current learning resource recommendation system,which make it difficult to effectively extract the recommended data relevance,a learning resource recommendation method based on graph comparison learning is proposed based on the research of lightweight accurate recommendation algorithm.Firstly,noise interference is applied to user nodes and learning resource nodes in the embedding layer to achieve augmentation of the original data;then graph encoders are used to obtain the attribute features of users and learning resources,and graph contrast learning is used to construct the auxiliary recommendation task.Finally,the main recommendation task and the auxiliary recommendation task are combined,thus alleviating data sparsity and improving the relevance of the recommendation data,so that more unexplored courses can be recommended.(3)Based on the proposed learning resource recommendation method,the block-chain based learning resource recommendation platform is designed and developed with the decentralized block-chain storage technology to address the problem that the current learning resource recommendation platform relies on the centralized data storage method,which leads to the leakage of user data.By designing a data uploading contract to upload key user data into the block-chain,it can avoid data being controlled by centralized storage methods,thus reducing the risk of data being maliciously tampered with or deleted.Finally,by embedding the uplink contract into the recommendation system and testing the core functions of the system,the results show that the platform has good usability and certain pervasiveness. |
