In order to research the effects of monetary policy to government bonds market, we adopt event study methodology to deeply analyze government bonds yield of four key terms, 1-year, 3-year, 7-year and 10-year, which construct from NSM term structure of interest rates model which use date on Shanghai exchange over the last year. Since we combine unit root with structural break test and t-test to fix on event period and adopt General Regression Neural Network (GRNN) model to forecast yields in every event period, our results are objective. The results show that the effects of adjusting the deposit reserve ratio to government bonds yields stranger than adjusting benchmark rate's when they were impacted by one same monetary policy. Our results have some important value to macro-control as well as to investors in China.