本节从路径长度入手，传统的直接限制路径长度会对部分长而有效的路径造成影

This section starts with the path length. The traditional direct restriction of path length will affect some long and effective paths. <br>In order to solve this problem, the concepts of k-step shortest path and k-step shortest element path are proposed; then according to the dynamic fixed length nature of k-step shortest element path, the definition of k-step pattern graph based on this concept is proposed, k-step mode The graph not only retains the original characteristics of the network model, but also greatly reduces the shortcomings of the huge scale of the network model. It can be used to partially replace the original network model and solve the problem of lack of practicality in the model-rich heterogeneous information network. <br><br>After solving the above key issues, we designed an automatic mining algorithm for complex heterogeneous information network paths based on reinforcement learning in Section 5.

This section starts with the length of the path, and the traditional direct limit path length affects some long and effective paths.<br>In order to solve this problem, the concept of k-step shortest path and k-step shortest metapath is proposed, and then according to the dynamic fixed length nature of k-step shortest metapath, the definition of k-step pattern diagram based on this concept is proposed, k-step pattern map not only preserves the original characteristics of network mode, but also greatly reduces the huge disadvantages of network mode scale, can be used to locally replace the original network mode, solve the problem of lack of practicality of network mode in the model-rich heterogeneous information network.<br><br>After solving the above key problems, we designed the automatic mining algorithm of complex heterogeneic information network path based on intensive learning in section 5.

This section starts from the path length. The traditional direct restriction of path length will affect the partial long and effective path.<br>In order to solve this problem, the concepts of k-step shortest path and k-step shortest meta-path are proposed. Then, according to the dynamic fixed length property of k-step shortest element path, the definition of k-step pattern graph based on this concept is proposed Step pattern diagram not only retains the original characteristics of network mode, but also greatly reduces the shortcomings of large-scale network mode. It can be used to partially replace the original network mode, and solve the problem of lack of practicability of network mode in heterogeneous information network with rich patterns.<br>After solving the above key problems, we design an automatic path mining algorithm based on reinforcement learning for complex heterogeneous information networks in Section 5.<br>