Abstract: The causal inference of transformable patent characteristics is the key to the cultivation, pre-evaluation and identification of high-value patents. The number of patents granted in China is increasing, but the number of transformable patents is small and the problem of low conversion rate still remains serious. This paper constructs a TOE framework that affects patent transferability from three dimensions, including technology(T), organization(O), and environment(E). Then we use LASSO and entropy method to screen and integrate multi-dimensional characteristics. Based on the TOE framework, the 16 features were synthesized according to the secondary dimensions of patent value, text features, organizational cooperation, organizational strength, and supply environment. Five comprehensive conditions for causal inference of transformable patent features were obtained. We combine NCA and fsQCA to explore the necessary conditions and key configurations of patent transferability, and reveal the driving modes and internal mechanism of the formation of transformable patents. fsQCA makes reasonable explanations for the complex reasons leading to specific results by dealing with the complex interaction between multiple antecedents. However, fsQCA can only identify whether a certain condition is a necessary condition for patent transformation. It cannot reflect the necessary degree of the condition for the result. NCA method can make up to its shortcomings. It can identify the necessary conditions while analyzing the necessary degree of their impact on the results. Meanwhile, NCA can be performed on each condition separately, so that the effect size of the necessary conditions is not affected by other conditions in the model. Therefore, we combine fsQCA and NCA to conduct causal inference research. Based on the empirical research on data in the field of fuel cells, we found that: (1) Patent value and organizational strength are necessary conditions for patent transformation, but the influence of necessity is small. It indicates that the formation of transformable patents is the result of complex interaction among multiple characteristic conditions, rather than a single characteristic condition. (2) The formation of transformable patents can be summarized into three driving modes: patent value and organizational strength hybrid driving mode, patent value and organizational cooperation hybrid driving mode, and organizational strength driving mode. The causal relationship between patent multi-dimensional characteristics and transferability has the characteristics of multiple concurrency and multiple solutions equivalent. (3) Based on the results of configuration analysis, we construct a CDEF mechanism model for the formation of transformable patents, that is: satisfying various technical requirements is the fundamental reason for achieving high patent transferability. The complex interaction of the three core conditions of high patent value, strong organizational strength, and multi-organizational cross-border cooperation, as well as the two auxiliary conditions of text characteristics and supply environment, constitutes diverse paths for the formation of transformable patents. On the one hand, this study reveals the internal mechanism and driving modes of the formation of transformable patents, which provides support for exploring the cultivation strategies of transformable patents. On the other hand, it expands the application of TOE framework in explaining "causal complexity", and provides an explanatory theoretical basis for the identification of transformable patents based on machine learning.