Multiple Relation Ranking Based On Heterogeneous Hypergraph

With the development of information technology,networks are playing more and more important role in real life.The network has made great progress in modeling and analyzing the problems of the systems.Many problems can be modeled and solved by methods the network.With the complexity of information in real society,the relationship between objects becomes more and more complex.The limitations of binary relationship network in complex relationship modeling are more and more obvious.At the same time,the simplification process of binary relationship network will also bring loss of characteristics and information.Therefore,researchers gradually realize the importance of multi-relationship modeling.Multiple relation ranking is one of the most important problems in multiple relation networks.Many practical problems can be transformed into multiple relation ranking problems to be solved,such as recommendation,prediction,evaluation of node influence and so on.In this paper,we selected two representative sub-problems,namely,scholars’ skill evaluation and disease gene prediction.Scholar skill ranking algorithm and disease gene prediction algorithm based on heterogeneous hypergraph ranking are proposed to achieve better results.We introduce the basic concepts,definitions and construction methods of hypergraphs and summarize the node ranking problems at first.Then we introduce the definitions and the topological structures of hypergraphs.Secondly,the definition of multiple relation and multiple relation ranking is given.And two multiple relation model are proposed.For the skill ranking problem,we analyze the factors affecting researchers’ skill ranking and propose a new model based on hypergraph theory to evaluate the scientific research skills.To validate the skill ranking model,we perform experiments on a constructed PLoS One dataset and compare the rank of researchers’ skills with their papers’ citation counts and h-index.We analyze the patterns about how researchers’ skill ranking increased over time.Our studies also show the change patterns of researchers between different skills.For the gene disease gene prediction problem,we propose a heterogeneous probabilistic hypergraph ranking method.In the heterogeneous probabilistic hypergraph,each node belongs to different hypergraphs in a probabilistic weight according to different edge types.We then carried out the experiment on a heterogeneous network constructed by integrating three data sets.The possible genes associated with disease were predicted by heterogeneous probabilistic hypergraph ranking method.The experimental results show that the heterogeneous probabilistic hypergraph ranking method achieves better results than the ordinary hypergraph ranking method and many classical disease gene prediction methods.