As the core bearing point of data in various business fields, human resource data is the key point to connect people, physical entities and business activities. By designing the label system of human data, abstracting the attributes of human data into graph structure, and realizing the automatic clustering of different attribute sets and the division of talent echelon through hierarchical clustering algorithm, the extraction and flexible combination of human data can be realized according to the dimensions of professional field direction, scientific research project undertaking, scientific research achievement acquisition and talent award. Lay the foundation for the training planning and target formulation of the talent echelon level.