基於狀態空間模型的區間震盪線日內交易策略以台灣加權股價指數期貨為例之交易實證

Abstract

對於時間序列的資料，如何估測其中的趨勢在許多學門都是一項重大議題。 在經濟和金融的領域中，趨勢可以作為一段時間內的波動累計的表現，在總體經濟學中擔任商業週期的分析工具，抑或是由於市場微結構交易機制造成的無法被觀測之合理價格。將一個時間序列萃取出其中趨勢確實是使我們瞭解這世界運作法則的一項關鍵，這類的問題在財金領域被稱作趨勢分解。在這篇論文，我們首先會回顧財金時間序列分析中關於趨勢分解的經典方法，並研究狀態空間模型相關統計特性。最後我們會對一些基於趨勢的日內交易策略在台灣股價加權指數期貨進行相關實證分析。

The problem of estimating underlying trends in noisy time series data has been a crucial issue in a variety of disciplines. The trend could be aggregate economic fluctuations, which is often referred to as business cycles in the macroeconomics literature. It could be the efficient price for the market, which is unobserved due to the microstructure noise that summarizes market trading mechanisms.Indeed, trend decomposition is one of the critical steps to understand the nature. In this thesis, we review classical trend decomposition methodologies and study the statistical performance of the state-space models.Also, a day trading strategy based on the state-space model is explored using Taiwan Capitalization Weighted Stock Index Futures.