In order to predict the remaining useful life (RUL) of rolling bearing accurately, a RUL prediction method of rolling bearing based on multiple criterions effectiveness analysis (MCEA), kernel principal component analysis (KPCA) and combined support vector regression machine (SVR) is proposed. For extracted feature, the effectiveness score of each criterion can be calculated, and the weight of each criterion can be determined adaptively, the feature will be sifted when the effectiveness total score of which is greater than its overall average, and the KPCA is used to remove the information redundancy among the sifted features, then the reduced feature matrix is established. The reduced features of multiple bearings are used respectively as the input of the SVR, the ratio p of the bearing running time to the whole life time, namely RUL are used as the output, multiple SVR models are established. And the selfadaptive method is used to determine the weight of each SVR model, the combined SVR prediction model can be established at last. Finally, the bearing which is different from the training process is used for testing, the reduced features are used as the input of the combined SVR prediction model, the p value of the bearing is predicted. The experimental results show that the RUL of the rolling bearing can be predicted accurately by the proposed method.