The invention relates to the field of flow
automation, and discloses a graph mining and graph distance-based flow recommendation method. The method specifically comprises the following steps: preprocessing, i.e., abstractly labeling an input flow set in the form of a
directed graph to obtain a flow sub-graph; discovering a pattern, i.e., decomposing a set output by the preprocessing step to obtain an upstream sub-graph, a candidate node set and confidence by a sub-graph mining and decomposing module, registering the upstream sub-graph, the candidate node set and the confidence as data entries in a pattern
list; recommending a flow, i.e., acquiring a reference flow by a recommendation module, comparing the reference flow with the upstream sub-graph in the pattern
list, selecting the most matched
data entry, and outputting the candidate node corresponding to the most matched
data entry as the recommended flow. The method has the advantages of high recommendation efficiency, smaller calculation complexity of the
algorithm, high recommendation accuracy, capability of supporting the
processing of a complex structure flow and higher application value.