A method and system for constructing a multi-mode graph index for weak structural correlation

Abstract

The invention discloses a method and system for constructing a multi-mode graph index oriented to weak structural correlation. The method is: 1) read the pattern diagrams in the pattern diagram collection of the target domain and generate a pattern diagram ID for each pattern diagram; 2) build a pattern diagram isomorphism tree: make a pairwise judgment on each pattern diagram, if There is a subgraph isomorphism relationship between the two schema graphs, then add a directed edge from the smaller schema graph to the larger schema graph, and get the schema graph isomorphism tree of the schema graph set; 3) For the schema graph the same Paper mulching performs frequent subgraph mining, finds common pattern graphs and adds them to the pattern graph isomorphism tree; 4) When there is a subpattern graph in the pattern graph isomorphism tree with multiple parent pattern graphs, it is the child pattern graph. The pattern graph retains the only parent pattern graph; 5) Calculate the minimum spanning tree of the isomorphic tree of the pattern graph, and perform depth-first traversal on it to obtain the optimal matching order of the pattern graph set. The present invention can greatly improve the matching efficiency.

Description

technical field [0001] The invention relates to a method and system for constructing a multi-mode graph index oriented to weak structural correlation, and belongs to the technical field of computer software. Background technique [0002] With the advent of the era of big data, massive multi-source heterogeneous data from the Internet and life is being generated and accumulated at an unprecedented speed, and there is a close correlation between these data. Data structure is very suitable for characterizing this kind of intrinsically related data. As an important means to realize efficient query on graph data, graph pattern matching is the basic technology of many graph-based data analysis and mining. [0003] Graph pattern matching refers to given a data graph and a specific pattern graph, and finding all matching results in the data graph with the same connection topology of nodes and edges of the pattern graph and the same attributes on the corresponding nodes and edges. W...

AI Technical Summary

Problems solved by technology

For example, the pattern tree algorithm, simply using the subgraph isomorphism method to construct the pattern graph index, has poor adaptability to the pattern atlas with weak correlation, and there are many patterns in the data atlas that have no isomorphic relationship but have common subgraphs Graph pairs, and the association between them cannot be established using only the pattern tree algorithm, this part of the calculation cannot be ignored

[0013] For the shortcomings of the pattern tree, the new improved algorithm utilizes the largest common subgraph between the data atlases, but the algorithm is more dependent on the effect of clustering, and needs to specify the clustering coefficient, which is less scalable for any data set , In addition, the pattern graph calculates the largest common subgraph in pairs, which consumes more time, and does not filter the common subgraph, and it is easy to perform unnecessary subgraph calculations during matching

Images