Large-scale Graph Mining Based On Deep Learning

With the gradual popularity of big data thinking and extensive study and application of deep learning,graph structures are increasingly used to characterize large-scale,intricate data in the real world,and in-depth mining of implicit information within the data gradually becomes the focus of research.In the era of information explosion,the traditional search engine based on keyword matching has been difficult to meet the needs of users to obtain information quickly,accurately and easily.The knowledge map can meet the new query needs by building semantic information entity graph.In this paper,we firstly provide a comprehensive survey on the development and construction of knowledge graph by reviewing and summarizing recent advances in the research and practice of knowledge graph systems in the relevant literature.In particular,our introduction includes the concept origin,development,and eventual formation of the knowledge graph,various data sources for the knowledge graph,the ontology construction and the entity extraction,and the process of knowledge mining,updating,and maintenance.Finally,we discuss the technical challenges,development trends,and future research directions of knowledge graph.Aiming at the problem of complex computation and data sparseness faced by large-scale data mining,this paper designs a deep learning representation algorithm based on word2 vec algorithm,which represents the graph node as low-dimensional vector and provides the possibility for mining graph data by using mature machine learning algorithms and linear algebra theory and tools.This algorithm uses the partial label information to guide the process of walking between nodes,and then uses the logistic regression classification model to classify the feature representation of nodes by multi-label.Compared with no label guide,the effect is improved obviously.In addition,this paper designs a combinatorial method to generate edge feature representation by using the vector representation of graph nodes obtained by network representation learning algorithm,constructs a classification model based on deep belief network,and realizes link prediction for complex network.