Abstract: Due to the application requirement of high layer atmosphere density model and ionosphere model, the forecast method research of 10.7 cm solar radio flux (F_10.7 index) is one of the important subjects. The middle term forecast of F_10.7 is more difficult than the long term and short term forecast because of the lack of comprehending physical progress and the laws of solar active region revolution. The main middle term forecast methods of F_10.7 are time-sequence model and empirical model in consideration of observational information of solar active region. In this paper, autoregression method was applied to the F_10.7 27-day forecast research. The period of forecast experiment was from May 1, 1996 to April 12, 2007. According to the results of forecast error analysis, the prediction precision was satisfying when solar activity was low and F10.7 showed obvious 27-day periodic tendency, but was unsatisfying when large region appeared on the disk or disappeared from the disk rapidly. Compared with the forecast result of American Air Force (AAF) forecast from September 21, 2005 to June 7, 2006, the prediction Accuracy of autoregression method was similar to that of AAF. From all of these, we can see this linear predictive method could reflect F_10.7 27-day periodicity, and was applicable for F_10.7 27-day forecast in a certain extent. The research result in this paper is a basis of the further work of constructing precise model of F_10.7 middle term forecast.