基本的短期利率擴散模型通常是由兩部份所構成：一是所謂的漂移項部份(drift term)，另一則是所謂的變異項部份(variance term)。在大部份的利率模型中，皆認定漂移項部份有所謂的均數復歸現象(mean-reverting)，在變異項部份當中，為了探討利率隨機變異性的特性，近來許多研究採用計量經濟上的一般化條件異質變異模型。然而，越來越多的研究發現，利率變異性對短期利率的變動卻是呈現不對稱性的反應。另外，最近的國外實證研究發現，在利率的均數復歸現象中也存在不對稱的現象，也就是說，當利率上漲或是下跌時，其後續向下調整或向上調整的力量與幅度未必是一樣的。由於過去的研究通常是在對稱模型的假設下進行，因此本研究希望重新在不對稱的架構中建立短期利率模型，再以國內的利率資料進行相關的實證探討，進而比較兩種不同架構下各個利率模型的績效。實證結果發現當模型納入不對稱漂移項和不對稱變異項的考量後，會比對稱的模型有較好的配適結果，對利率變異性也有較好的預測能力。

A basic interest rate diffusion model is comprised of a drift term and a variance term. One particular feature in the drift of most interest rate models is that interest rates appear to be pulled back to some long-run average level over time. This phenomenon is known as “mean reversion”. On the other hand, many stochastic volatility models combined with generalized conditional heteroskedasticity of volatility processes were developed. Those works often assume the conditional variance as a symmetric function of unexpected shocks to short-term interest rates. However, more and more studies found asymmetric volatility responses to the interest rate shocks. In addition, some studies reveal that interest rate models with asymmetric drift terms are more appropriate in explaining mean reverting phenomena. In other words, the drift in the interest rate model is different when the interest rate increases or decreases. Therefore, new interest rate diffusion models with asymmetric frameworks are developed to conduct an empirical study by using Taiwanese interest rate data. The result shows that diffusion models with asymmetric drift terms and asymmetric volatility are bettern than diffusion models with symmetric ones about model fitting and forecasting the future volatility of interest rate changes.