Determining Optimum Order of Autoregressive Model and the Application to Long-range Forecast

The methods of determining the optimum order of autoregressive (AR) models, such as FPE, AIC, BIC, L1and L2 were summarized and tested by using the monthly mean temperature data in Qingdao. The selected orders of the AR model by use of the FPE, AIC and L1 criteria are the highest, medium by L2 and the lowest by BIC, respectively. Additionally, a recurrence method of AR model was suggested to forecast the monthly mean temperatures in Qingdao. It has been proved by the forecast practice that the low order AR model from the BIC criterion is more efficient