Survey of link prediction method in heterogeneous information network
Link prediction is the prediction of unknown or future links based on known information in the network,and is one of the research hotspots in the field of data mining.Heterogeneous information network can accurately portray the semantic information from data and improve the efficiency of downstream data mining tasks.Therefore,link prediction method on heterogeneous information network needs to take into account the topological characteristics and semantic characteristics of the network,which brings new challenges to the link prediction task.On the basis of previous research,this paper systematically sorts out the link prediction methods on heterogeneous information network in the past decade.Firstly,the concepts of heterogeneous information network and link prediction are introduced.Secondly,the link prediction methods in heterogeneous information network is classified,and the link prediction methods in different types of heterogeneous information network are summarized.Further more,the typical methods are introduced in detail.Then,the application of link prediction methods in heterogeneous information network are sorted.Finally,the problems that need to be addressed in further research in this field and potential future resarch directions are summaried.
heterogeneous information networklink predictionmeta-pathsupervised learning
NETL
NSTL
万方数据
