Feature selection can eliminate the redundant features in the original features set, find an optimal subset of features and enhance the classification accuracy and efficiency in machine fault diagnosis. A feature selection method based on evolutionary Monte Carlo is proposed. Support vector machine (SVM) is applied as the fault classifier, the evaluation criterion is the Wrapper model, and the evolutionary Monte Carlo is implemented for optimal feature subset selection. This method is applied on the feature selection of the rolling bearing fault diagnosis based on vibration signal. Experimental results indicate the proposed method is effective for feature selection in fault diagnosis.