In order to accurately predict market interest rates，a novel Ｒeverse Ｒestricted MIDAS ( RR-MI-DAS) model is developed on the basis of the MIDAS and ＲU-MIDAS models． The ＲＲ-MIDAS model can be applied to the prediction of high frequency variables using low frequency variables when the frequency mismatch is pretty large． SHIBOＲ is used as a representative of market interest rates，and an empirical analysis of SHIBOＲ forecasts is conducted． The empirical results show that the ＲＲ-MIDAS model outperforms the others in terms of goodness of fit and prediction ability since it is able to explore the dynamic relationships among var iables． The results show that both macroeconomic variables and the capital market information could influence the money supply and demand in one week，or even one day，and will quickly lead to a change of SHIBOＲ．Moreover，robustness tests are implemented to illustrate the efficacy of the ＲＲ-MIDAS model and the reliability of the empirical conclusions．