本研究主要針對改善利率期限結構平滑度之兩種roughness penalty方法，比較其配適能力優劣。本文在Fisher，Nychka, and Zervos（1995）所提出之smoothedsplines 加上roughness penalty模型的基礎上，首先採用generalized cross-validation（GCV）模型，來估計台灣政府公債市場的遠期利率期限結構，實證結果發現GCV模型對於台灣政府公債市場利率期限結構的估計，在平滑度上具有較佳之配適結果表現。本文另外採用Waggoner（1997）所提出之Variable roughness pena1ty（VRP）模型進行比較，結果發現VRP 模型在精確度配適能力較GCV 模型為佳，但在平滑度方面之配適能力則略遜一籌。

　　This paper investigates the fitting performance of term structure of forward rates based on the roughness penalty correction model. At first, the generalized cross-validation (GCV) model proposed by Fisher, Nychka, and Zervos (1995) is used to estimate the term structure of forward rate in Taiwan Government bonds market. The empirical results reveal that the GCV model can produce a better performance in fitting smoothness. In addition, comparing the results with Variable roughness penalty (VRP) derived by Waggoner (1997), the VRP model outperformed the GCV model in fitting accuracy but an inferior results in smoothness.