利用資料分析進行金融市場之探討與研究

Abstract

了解金融市場行為是非常有趣的問題與題目。隨著科技技術的發展，目前的環境已經擁有大量的數據供我們參考和研究。這些數據訊息的產生，讓我們可以有跡可循的去解決問題。大數據分析也讓我們判斷的更加精確，進而解決各領域的問題。當然，這些問題也包含了金融交易領域的問題。數據分析提供了一種科學方法來判斷和驗證數據背後所帶給我們的線索，提高了生產效率也讓我們有更多的創新機會。本文旨在通過數據分析探索金融市場的行為。在此論文，我們透過對市場行為的了解提出了兩種交易策略。這兩種交易策略分別是時間突破區間策略（TORB）以及持續行為策略（PBS）。我們還藉由數據分析驗證了這兩種策略背後的金融市場行為。對於這兩種策略，我們獲得了最佳的結果分別是，在TORB中，在p值為3.1×10-5的年收益率為20.28％；在PBS中，在p值小於1×10-5的年收益率為27％。我們也針對TAIEX期貨交易數據進行了實驗來分析這兩種策略的信號與交易者行為之間的關係。我們發現這兩種策略的交易訊號與法人的投資行為呈顯著性的正相關。而這個正相關在國外投資法人更為顯著。

Understanding the market behavior is an interesting issue in financial markets. With the development of information technology, a considerable amount of data has been generated and thus can be utilized to produce useful information and insights to better solve various problems in many fields including finance. Data analysis provides a scientific way to capture and verify the information behind the data, which promotes productivity, improves efficiency, and supports innovation. In this dissertation, we aim to explore the financial market behavior via data analysis. Thus, we proposed two trading strategies, which are timely open range breakout (TORB) and persistent behavior strategy (PBS). We also explored financial market behavior behind these two strategies through data analysis. For these two strategies, we obtain the best performance, 20.28% annual returns at a p-value of 3.1×10-5, in the TORB, and 27% net annual returns with p-value less than 1 ×10-5, in PBS. We conduct experiments on a TAIEX futures transaction dataset to analyze the relationship between these two strategies’ signals and traders’ behaviors, and find the signals are in the same direction as institutional traders, especially foreign investment institutions.