Recommendation of fast learning resources based on knowledge map in storm distributed computing framework
Since the online learning resource recommendation holds low accuracy or poor real-time performance,this paper presents the knowledge representation of users and resources using knowledge graphs,and uses the long-short time memory network to optimize user resource feature differences.Thus,the resources with the smallest feature differences with the users can be pushed to the users.First,a sample of online learning records is obtained,thus the knowledge representation of entity feature relationships is performed using the knowledge graph,and the head and tail entity and relationship feature vectors in the knowledge graph are generated with the help of the Storm distributed framework.Second,the minimum feature difference objective function of user-resource entities is established,and the minimum feature difference objective function is optimized by using long and short time memory networks.The parameters of the long-short time memory network are again solved by the Storm distributed platform,so as to quickly generate a stable and relevant resource recommendation model.The experimental results show that the online resource recommendation using the knowledge graph and the long-short time memory network under the distributed framework of Storm can achieve high accuracy and operational efficiency,and has strong adaptability in dealing with real-time recommendation of large-scale resources.
resource recommendationknowledge graphStorm frameworklong and short term memoryTransD model
国家科技期刊平台
NSTL
万方数据
