With the current data collection capacity and storage capacity greatly enhanced, the importance of largescale data mining analysis more and more important. However, the analysis of largescale data mining is not an easy thing. Therefore, in order to be able to more efficient analysis of these data, many new algorithms and data structures are gradually introduced to the data mining analysis. This paper is based on the correlation analysis, based on this article, proposed a called advanced frequent pattern mining (AFPM) algorithm. This algorithm is based on the prefrequent pattern tree (PFPtree) to improve the performance of association analysis and provide the corresponding algorithm to implement the association analysis based on this data structure. It is proved that this new data structure is superior to FPgrowth algorithm in association analysis problem through a large number of experimental data.