鑒於近年財務論著主要使用的BPN神經網路模型有其限制，本研究希冀ぇ比較VECM 、BPN，暨 RN 對國內公債價格之預測績效。 え找出景氣和預測變數關係，探討各個時期各 計量和神經網路模型是否有互補性或替代性，預測績效是否受經濟環境影響。影響殖利率的 因素可拆解成實質利率、預期物價上漲率和風險貼水三層面，我們亦循此三項範疇選取變數 以求周延。研究貢獻主為：ぇ引進 RN 神經網路，驗證比較其與以往財務測領域慣用之 BPN 表現差異。並比較各模型在不同景氣狀況下，對不同期長債券預測力差異，冀探究工具之選 擇、應用與搭配。え分析總體變數、風險變數對殖利率影響，俾進一步瞭解影響債價的相關 因素。

　　　　　In this study, we empirically explore the relative predictive abilities of Vector Error Correction Model, which serves as a representative econometric model, Back Propagation neural Networks (BPN), which is adopted by most current studies in the application of neural networks in finance, and Reasoning Neural Networks (RN) for Taiwan's government bond prices during 1992-95. We also examine the extent to which the relative predictive abilities vary in different phases of economic cycle, investigating if the models substitute or complement one another. Our explanatory variables include all potential drives to real risk-free rate, expected inflation rate, and risks premiums. This study contributes to the concurrent literature in two aspects: (1)Few, if any, prior study explores whether and how various neural networks and/or econometric models perform under different macro-economic variables. Our empirical results may indicate an appropriate forecasting model (models). (2)BPN, the prevailing model in financial forecasts, is subject to a few short- comings and may thus be a sub-optimal model. This study analyzes if RN is more effective in forecasting bond yields than BPN.