Research On Algorithms For Graph Classification Based On Closegraph

With the development of computer science and information technology,data mining technology has been widely applied to artificial intelligence,pattern recognition,bioinformatics and many other fields.The demand of ruining on complex data is rising now. Experts have paid attention to these fields and tried to solve the problems by virtue of the experience of structured data mining. In this paper, we do the research on bases on graph date mining problem.At present,the urgent problem in the field of the mining graphs is how to improve the subgraphs mining algorithm efficiency.Due to the frequent subgraphs mining generate a great results set,this restricts the algorithm performance to some extent,and mining maximal frequent subgraphs can effectively reduce the results set of the frequent subgraphs.Therefore,this paper focuses on the research to mining frequent subgraphs algorithm and graph classification algorithm bases on closegraph, which is improve the efficiency of the algorithm. In allusion to these problems, the paper dose some reach as follows.Firstly, this paper proposes a new algorithm BPCG which based on the study of the typical mining frequent closegraph pattern mining method. A new storage structure of the frequent subgraph is used for the algorithm, which can be directly to the expansion of the most frequent adjacent edge and calculate the threshold of support without scanning the database, and then to narrow the search space and reduce the unnecessary operation, brother pruning strategy and delete local frequent edge are adopted in the algorithm also.Secondly, baesd on the improved algoriht for mining frequent closegraph, with frequent closegraph mining results as a set of features candidate, a graph classification algorithm CGC has been proposed in this paper. In this algorithm the way how to get graph features and how to construct classifier have also been introduced.Finally, the example and experiment also been proposed to give instruction and validation for this BPCG algorithm. The runtime of this algorithm has been significantly improved when mining large graph set. At the end of this paper the efficiency and accuracy of CGC algorithms have been tested by experiments...