Abstract: The observation of time series may be influenced by outliers.If we forecasts directly,neglecting the influence,will lead to false result.So,it is important to find proper outliers detection method.Based on the theory of Bayesian statistical inference,in this paper we put forward Bayesian method of positioning outliers in autoregressive model firstly;and then,under the condition of normal-gamma prior information,we put forward computation method of posterior probability based on mean shift model and variance inflation model respectively,and estimate the outliers with Bayesian method;at last,the method is applied in the research on the data modeling of ionospheric VTEC series,compared with the forecasting results on unmodified and modified to test the efficiency.