This paper presented a prediction method based on the regression sliding average (ARMA) model to achieve the vibrational data prediction of rolling bearings. Firstly, the early failure data of the full life cycle for the rolling bearings were selected as samples to calculate the autocorrelation coefficient and the deviation corre- lation coefficient of its truncated sequence. Then, the minimum information criterion (AIC) was adopted to order the ARMA model. By using the least square estimation parameter, the ARMA model was established. And the bearing data under the same and similar working conditions were input to the established ARMA model. The comparison between prediction data and the actual failure data was made to calculate the prediction accuracy. The research results show that this method can accurately predict the actual state of the bearing and the predic- tion result of the same working condition is better than that of the similar working condition.