Abstract: In the current era of stock-based urban redevelopment, large amounts of inefficient urban land waste scarce land resources. Scientifically identifying, planning and governing mixed-function inefficient land has therefore become a major issue in urban planning research and practice. From a "human-landindustry" synergy perspective, this study clarifies the connotation of inefficient urban land identification and proposes a step-by-step intelligent identification method. The first round, preliminary identification based on hierarchical feature combinations, reveals the spatial-location hierarchy of inefficient urban land. The second round, optimized identification based on multidimensional feature clustering, identifies four types of "human-land-industry" feature clusters. The third round, fine-grained identification based on graph-structure deep learning, uncovers contradictions and conflicts in the allocation of resource elements within inefficient land units. The results show that step-by-step intelligent identification can finely identify and deeply explain the characteristics and mechanisms of inefficient land, providing a new pathway for refined governance and planning response in the new era.