Abstract:Under the macro background that many core technologies in China are facing the "neck sticking" problem, fully mining and discovering the comprehensive information of innovation preferences of different enterprise groups will help to further improve the efficiency of enterprise innovation resource allocation, and maximize the innovation ability of enterprises and the efficiency of government investment. In order to accurately capture various types of enterprise innovation preference information, this paper proposes a patent data-driven group portrait model based on Fuzzy three branch concept lattice for enterprise innovation preference. Taking patent IPC information as the breakthrough point, it achieves the goal of leadership, stability, potential and innovation The differentiation and multi granularity characterization of innovation preferences of four types of dangerous enterprises. On this basis, the innovation partial fuzzy three branch concept lattice with similarity is constructed to reveal the relationship of innovation preference among enterprises. Taking the data of China's IC design industry as an example to verify the effectiveness of the model, the results show that the model can clearly show the field layout and structural layout of various types of enterprises' differentiated innovation preference, and provide theoretical basis and reference for enterprises' competitive strategy formulation and government innovation decision-making.