Graph data processing method for separately processing low-degree points and core sub-graph

Abstract

The invention discloses a graph data processing method for separately processing low-degree points and a core sub-graph. The invention specifically discloses a method for marking graph nodes, a methodfor storing and arranging graph data, and a graph data processing method. The method for marking the graph nodes comprises the following steps of: marking nodes in a graph as a core graph node type and an edge tree node type, wherein the core graph node type comprises a core graph internal node type and a core graph edge node type, the edge tree node type comprises an edge tree internal node type, an edge leaf sub-node type and an isolated point type; then performing storage and layout by utilizing the marked node type, specifically, adopting a CSR ROW array and a CSR COL array in a CSR storage structure to store all the edges in the graph, and adopting an edgelist array to store the edges in the edge tree sub-graph; and finally, processing the CSR storage structure and the edgelist arrayobtained by the layout by adopting the method, and processing the edges in the core graph based on the CSR storage structure and processing the edges in the edge tree sub-graph based on the edgelistarray.

Description

technical field [0001] The invention relates to the field of graph data processing, in particular to a graph data processing method for separately processing low-degree points and core subgraphs. Background technique [0002] Graphs are extremely abstract and flexible, and they can fully express the connections and dependencies of things in nature. Many problems can be efficiently solved with the help of graph-related algorithms under the support of graph theory, such as graph coloring, network routing, network flow, etc. In addition, graph data processing can be used to mine and analyze massive, sparse, and hyperdimensional associations, and has been widely used in social networks, transportation networks, biological information networks, knowledge graphs, and GNNs. Breadth First Search (BFS) graph data processing method is a basic graph data processing algorithm. Many graph processing methods can be built on the basis of BFS graph data processing methods, such as PageRank...

AI Technical Summary

Problems solved by technology

[0003] Due to the nature of power graphs in natural graphs and social networks, the number of high-degree points is small and the number of low-degree points is large, which is significantly unbalanced. The existing methods mainly focus on the processing of high-degree points, but the low-degree points The point will also bring a lot of random access memory

The existing technology does not consider the particularity of low-degree points, and processes all points uniformly without distinction. It does not realize that low-degree points will also cause serious random access memory, which seriously restricts the performance of the graph computing system. One of the main reasons for the inefficient performance of some graph processing and graph computing framework solutions

Images