Graph mining based on co-occurrence frequent-item tree and inverted matrix
Li Ta · Jisuanji yingyong yanjiu · 2014
For the problem of subgraph isomorphism detection must be avoid in graph mining in the era of big data,this paper adopted the social network information propagation model to compute the association strength of node label,and then ran the joint probability distribution to calculate the probability support of the node label set. It adopted the improved inverse matrix +COFI-tree mining algorithm to perform the frequent itemsets mining in transaction data set consisted by node label according to the probability support. Comparative analysis of experimental results show that the proposed method is faster than the frequent subgraph mining algorithm in a single large graph,the returned results are more than frequent subgraph mining algorithm,and can find some interesting patterns that can not be found by the traditional frequent subgraph mining algorithm. And it has high memory utilization efficiency and can support large data mining,so it surpasses the FP-tree-based mining algorithm.