文章对两因子仿射模型在银行间国债市场的预测能力进行了实证检验,结果表明,模型对于3-4个月内剩余期限在7年期以下的品种预测误差较小,可信度较高。利用该预测方法,构建了基于极大似然框架的预测效果指标,证实了模型参数的稳定性较好。在归因分析的基础上,又构建了自适应的模型预测误差修正方法,这一方法对于预测期在4个月以上的中短期国债的预测误差修正效果最为显著。 The paper tests the predictive ability of the two-factor affine model in the inter-bank bond market. The results show that the model has less error and higher reliability for breeds with residual maturity less than 7 years within 3-4 months . The prediction method based on the maximum likelihood framework is used to establish the forecasting performance index, which proves the stability of the model parameters is good. On the basis of the attribution analysis, an adaptive model prediction error correction method is also constructed. This method has the most significant effect on the prediction error correction of short and medium-term government bonds with a forecast period of more than 4 months.