This study assessed the usefulness of a rice yield prediction model using fixed soil nitrogen mineralization parameters as input variables. When the soil nitrogen mineralization parameters obtained 2016 in the paddy field, yield predictions of cv. Koshihikari in same paddy field applied rice straw and fast-acting fertilizer every year were evaluated from 2017 to 2020. Four levels of primary fertilizer and four levels of supplemental fertilizer were applied, the actual and predicted yields were compared. Results revealed that the yield prediction error averaged at 53 kg 10a-1 for RMSE and 9.1% for MAPE. These findings indicate that previously obtained mineralization parameters are useful for yield prediction over 5 years.