A method for solving the uncertainty of support vector hyperparameter (SVP) selection in rotating machinery fault diagnosis, named generalized patter n search (GPS), was presented in this paper. Firstly, the range of parameters wa s determined by SVP with some prior information. Secondly, optimum parameters we re obtained through a global search and a local search. Finally, an improved sup port vector machines(SVM) model was introduced into the rotating machinery field for fault diagnosis. The experimental results show that the search efficiency, the optimizing parameter performance and the fault diagnosis accuracy were significantly improved by the method. This method has a better application prospect in the rotating machinery fault diagnosis field.