In order to improve the low efficiency of data mining with high density and complex heterogeneous data network, a data mining algorithm based on multi dimension association structure is proposed. Firstly, on the basis of the data personality characteristics and the differences of the data storage, forwarding and processing in the heterogeneous large data network, the data definition and multi-dimensional correlation model of the heterogeneous data network are given. Then, based on the large data network, the paper proposes a multidimensional association fine-grained data mining algorithm based on the reconstruction of the large data units, the dimension replacement, the granularity and the granularity. Finally, the efficiency of the algorithm is compared with the coarse grained algorithm and the linear structured data mining algorithm. The experimental results show that the proposed algorithm has better performance.