Abstract: The classical single kernelized rough set model ignores the interaction between different kernelized relations. In order to solve this problem, this paper is devoted to the construction of the fuzzy rough set model based on multi-kernelized granulation. In this study, the optimistic and pessimistic rough set model, which is derived from a family of the kernelized relations, is deeply explored to multi-kernelized granulation space by defining the S-T multi-kernelized lower and upper approximation operators. Finally, we apply these measures to evaluate and select features of classification problems. The experimental results verify the interaction in different granulating relations.