A Fast Search Method for Entity Correspondence Between Similar Subgraphs

Abstract

The invention discloses a method for quickly searching entity corresponding relations between similar sub-graphs. The method comprises the following steps of S1, carrying out vectorization expressionon network data; 2, calculating a matrix of vector distances between entities; 3, calculating a series of entities which are closest to each entity; 4, searching for a similar ordered set similar to each entity in query sub-graphs given by a user; 5, obtaining a similar sub-graph set of the given query sub-graphs; 6, obtaining an ordered similar sub-graph set; 7, clustering the entities in the query sub-graphs given by the user; 8, for each entity in each similar sub-graph in the ordered similar sub-graph set, calculating average vector distances of all the entities in each cluster and the cluster with the minimum average vector distance; and 9, performing calculation to obtain the corresponding relations between the entities in the similar sub-graphs and the entities in the query sub-graphs given by the user.

Description

technical field [0001] The invention relates to the technical field of quick search for similar subgraphs, in particular to a quick search method for entity correspondence between similar subgraphs. Background technique [0002] At present, network data is widely used to describe the relationship between entities, such as the social relationship between people in sociology, the interaction between proteins in biology, and the transaction relationship between financial companies. The scale of network data will greatly affect the efficiency and accuracy of users' network analysis. How to search for similar subgraphs of a given subgraph in large-scale network data, such as mining similar transaction patterns, social relations, etc., has become a hot topic in the fields of the Internet, biology, and business. [0003] For network data, each entity can be, person, place, item, etc., and the relationship between entities can be personal relationship, email communication relations...

AI Technical Summary

Problems solved by technology

The method of finding similar subgraphs in huge network data has the problem of taking too long. Many existing methods search based on the category of each entity and relationship, because it comes with its own classification. Greatly reduces the search space

However, in network data without categories, it is impossible to effectively search for similar subgraphs for a huge amount of data, such as VF3, L2G, LAD, RI and other methods. The RI algorithm is to add some simple methods to the deep backtracking search. The rules are pruned and optimized, and the simple rules mean that the time to evaluate whether the rules are met is short; the VF3 algorithm also defines some rules to optimize in the deep backtracking search, and it has been proved that the speed is faster than other RI, L2G and other algorithms

However, these algorithms are difficult to complete the search for similar subgraphs on the entire network structure in the interaction-level time requirement

[0004] In addition, the prior art lacks the ability to finally find out the corresponding relationship between the points in each similar subgraph and the points in the original image. The corresponding relationship between points can help users further analyze and explore similar subgraphs, such as finding similar subgraphs. Center points, outliers, etc. in the graph

Images