Research And Implementation Of Graph Pattern Mining Based On Graph Summarization

Although graph pattern mining is a classic field of data mining.However,with the increasing scale of data,the scale of graph data is larger and the graph structure is more complex.The information hidden in the graph becomes more difficult to be excavated.Therefore,graph pattern mining faces many challenges at present.For example,in large graph data processing,the traditional pattern mining algorithm increases exponentially with the decrease of support,which makes it difficult to complete the mining task in a reasonable time.Frequent subgraph mining algorithms usually produce a large number of frequent subgraphs with even exponential numbers,which seriously affect the results of mining.The research of graph abstract is mainly aimed at the simplified operation of large scale graph operation.In recent years,because of the explosive growth of data scale,the technology of graph Abstract field is developing at a high speed in recent years,and has made great achievements.The graph abstracts are abstracted using the method of graph summarization,and then the graph pattern mining is carried out on the summary diagram,which can effectively avoid producing too many output graphs,and its time and space complexity is much smaller than that of other graph mining algorithms because of the characteristics of the graph summary.Therefore,this paper focuses on the combination of graph summarization algorithm and graph pattern mining algorithm.The main work is as follows:(1)frequent subgraph mining based on graph summaries.The advantage of graph summary algorithm in accelerating algorithm processing is introduced into the field of frequent subgraph mining,and Pattern Discovery by Dense Graph Summarization(PDDGS)algorithm based on graph digest is proposed.The PDDGS algorithm uses graph summarization algorithm to partition the graph,and then subdivides the result into input to carry out frequent subgraph mining.Since the algorithm is introduced,the graph is divided and then excavated to reduce the scale of the graph.The time complexity of the algorithm is reduced in the reasonable loss of the mining results.Experiments verify the feasibility and effectiveness of applying graph summarization algorithm to frequent subgraph mining.(2)interesting subgraphs based on graph summarization algorithms are found.On the basis of the algorithm of graph summary,the concept of word label is introduced,and the Dense Graph Summarization of Vocabulary(DGSoV)algorithm with word label is proposed.The DGSoV algorithm first abstracts the graph,and gives the word label to each of the sub icons produced by the summary,and uses the word label to describe the graph structure;then,on the basis of the structure,it interpretates the mining pattern,and successfully excavates the realistic model.The experimental results show that the advantages of the graph digestalgorithm in processing time are preserved while mining interesting patterns.The effective combination of the graph summarizationalgorithm and the graph pattern mining algorithm makes the processing efficiency improved when the improved algorithm results within a reasonable error range.At the same time,this method also provides a new graph mining method,which enriches existing graph mining algorithms and has certain practical significance and application value.