The invention relates to a distributed large graph division method based on affinity clustering, which mainly aims at unweighted
undirected graph division
processing, initially divides a graph into specified k sub-graphs based on Boruvka
algorithm hierarchical affinity equilibrium graph clustering, and iteratively combines two types of vertexes with closer distance by taking vertex similarity as
distance measurement, so as to obtain k sub-graphs. The vertex with the minimum adjacent point similarity and value in the sub-graph is removed to constrain the sub-graph with the overlarge scale, and the division quality is close to that of an existing large graph division method under the condition that follow-up optimization is not carried out; in order to solve the problem of edge
cutting rate optimization between large-scale sub-graphs, dimension reduction operation is designed, an initial division result is mapped into a vertex sequence, the vertex sequence is divided into a certain number of sub-slices, two sub-slices in adjacent sub-graphs are randomly selected, vertexes are migrated according to mutual exchange positive revenue and single-point positive revenue, and the edge
cutting rate of the large-scale sub-graphs is optimized. Therefore, the purpose of optimizing the edge
cutting rate is achieved.