探討外國機構投資者未平倉量對台指期之影響

Abstract

本研究旨在分析外資法人在現貨市場、期貨市場、以及選擇權市場的策略佈局，其中更針對未平倉量為標的，探討外資在量大時的行為表現。為探討外資籌碼對於台指期未來走勢的影響，並藉由這些影響，作為預測未來股市的依據，提供投資人作為參考。 本研究利用倒傳遞類神經網路強大的非線性搜尋能力，學習外資法人在現貨、期貨、選擇權佈局的行為模式，並且結合金融工程物理學，透過將變數做一階與二階變量的處理，發掘其潛在隱藏的知識。 本研究針對外資法人的未平倉量作分析，以不同的輸入變數及門檻建立十二個實驗組分別學習以及預測台指期貨未來一日、三日及五日的漲跌，以過濾出外資法人籌碼在量大時期的資訊。最後透過交易策略的設計檢驗績效。 研究結果顯示，門檻的篩選的確具有去蕪存菁的效果，外資未平倉量在量大時具有較明顯的行為模式，得以提高類神經對於台指期的預測能力。

This study aims to analyze strategies that the foreign capital takes on the targets of spot market, futures and options. Particularly focusing on the issue of the open interest, this study explores the behavior of the outstanding amount of open interest. Further, to explore the impact of the foreign capital on the future trend of the TAIEX. Basing on this analysis influences, this study predicts the future price of the TAIEX. The results could provide information as a reference for the investors. This study takes the advantage of the nonlinear search capabilities of back-propagation neural network (BPNN), combining with the concepts of econophysics from Chen and Hsu (2010), to explore its hidden knowledge and to learn foreign capitals’ pattern behavior of the investment in stock, futures, and options. This study analyze the foreign capitals’ open interest, and build ten different experimental groups according to different input parameters and thresholds. The experiment is to predict the rises or falls of the TAIEX for the next 1, 3 5 days, to filter out the information of the outstanding amount of open interest. In the end, the study tests the performance by implementing the designed trading strategy. The result shows that the threshold does have the effect of improving the BPNN’s predictive ability for TAIFX, since the outstanding amount of interest presents more obvious pattern behavior with the filter.