探討匯率波動的原因是國際金融的重要主題之一，許多研究驗證未預期經濟指標對長期平均匯率影響，然而匯率短天期變動則相對缺少討論。觀察匯率的長期日平均變動較為穩定，但在重要經濟指標公佈時，受未預期訊息的影響，匯率可能產生短天期反轉。所謂反轉是指和原本波動趨勢不同出現較巨大由漲勢轉跌勢或由跌勢轉漲勢的短期狀況。本文修正Caginalp and Laurent (1998) 價格K 線型態指標研究中反轉定義擴大為移動性5 天期的反轉。未預期之總體經濟指標是指官方實際公佈值與市場共識值的差異，此差異值可能必須超過或低於某特定數值才會影響短期反轉之機率。以決策樹演算法作為第一階段門檻值篩選工具，過濾個別未預期消息面指標與反轉方向互動的關係。第二階段則輸入各門檻值並建構羅吉斯迴歸模型。選擇全球金融市場主要交易貨幣並將資料分為次貸風暴前後兩期間探討未預期總體指標對匯率反轉的影響。本研究主要結論：(1) 各國造成匯率短天期反轉的關鍵因素不一，但同一貨幣在兩階段受到指標訊息影響因素相當一致。(2) 未預期經濟成長類基本面指標訊息在次貸前後皆為影響匯率反轉的因素。(3) 其中五種指標之未預期消息面必須注意門檻值（非0）影響。

Understanding fluctuations in foreign exchange rates is one of the major issues of concern in the international finance market, as the trend of a foreign exchange rate may reverse when announced macroeconomic indices do not match forecasts. Mussa (1976) proposed that about 90% of fluctuations in exchange rates are due to unexpected news. Therefore, this study presents the two-stage TREE/LOGIT model to identify which indices with thresholds lead to such short-period reversals. The first stage applies a decision TREE algorithm to determine the possible threshold difference between the forecast index and the announced one that correlates to the reversal. The next stage makes use of the LOGIT model with thresholds to quantify the statistical significance of a combined set of indices as the optimal predictor of a reversal. With this model we analyze each of the five major currencies exchanges rates, for before and after the u.s. subprime mortgage crisis in 2007, and identify the relevant indices that precede a reversal under different investment risk aversions. We conclude that the category of economic growth indices is the primary factor for predicting an exchange rate reversal, either before or after the crisis, for all the currencies in our study. Each currency in turn has its own other different decisive categories before and after the crisis with various trading suggestions under significant threshold news.