Research On Heterogeneous Information Network Similarity Search Technology Based On User Feedback

With the rapid development of information technology,a variety of information objects from different aspects,different levels,in different ways to interact with each other,constitute a large and complex heterogeneous information network in real life.We can discover the implicit knowledge behind the heterogeneous information network by analyzing and mining the complex information network.As heterogeneous information networks contain rich information,the high value of implied knowledge,similarity search on heterogeneous information network has become a hot topic for many scholars at home and abroad.This paper base on the research status of information network and analyze the similarity search technique based on meta-path and the clustering technology of mixed attribute data.Similarity search on heterogeneous information network takes as the research goal.This paper also focuses on the semantic information expressed by link-path of similarity search technology and clustering method of feature-based similarity search.Combine user feedback,link path,semantic information,classification attribute clustering,etc.to provide accurate similarity search results.The main works as follows:First,the similarity search algorithm based on user feedback is proposed.As the similarity method based on link path has the disadvantage of ignoring the object attribute and feature-based similarity calculation makes full use of the object's attribute characteristics but ignores the connection between objects.Therefore,we combine the advantages of the two methods,user feedback,and propose the the similarity search algorithm based on user feedback.It makes full use of the link path and object attribute characteristics to improve the accuracy of similarity search and user's satisfaction.Secondly,the similarity search algorithm of path enhancement is proposed.As the similarity search algorithm based on user feedback does not fully reflect the user's intention,that is,the path selected by the user contains some of the attribute characteristics of the object,reflects the user bias on some of the characteristics of the property indirectly.We propose the similarity search algorithm of path enhancement,which makes full use of user feedback expressing user intention.The final similarity search results are more biased the user's search intent.It further improve the accuracy of similarity search results and customer satisfaction.Finally,on the DBLP dataset,the two similarity search algorithms proposed in this paper are validated by experiments.Through the analysis and comparison of the experimental results,the two similarity search algorithms proposed in this paper have further improved the quality evaluation index nDCG and improved the user satisfaction comparing with similarity search algorithm based on link path.