Abstract: A method for predicting generalized time-varying ARMA series is proposed.The prediction formulas of time-varying and generalized time-varying series and their mean squared errors are presented.The method can analyze the generalized time-varying series(or signal) whose mean,variance,autoregressive and moving average coefficients vary with time.It can be used in meteorology,communication,automatic control,structure response analysis,fault diagnosis and other fields.The precision of the presented method is higher than the traditional method in predicting generalized time-varying ARMA series.