| This study intends to empirically test the market efficiency of the TEI and its market cointegration in terms of predictabilities among the rate of return of the TEI, the Taiwan Stock Exchange Weighted Index (TAIEX), and four major U.S. indices: the NASDAQ Computer Index, the NASDAQ Index, the Dow Jones Industrials Index and the Philadelphia Semiconductor Sector Index. The term "predictabilities" refers to how the possibilities offered by the information transformation of international markets in different time zones may provide superior profits from equity trading.;This study used the NeuNet Pro software package to establish an artificial neural network market prediction system based on back-propagation algorithms embedded as neurorules. This system accepts trading information from major Taiwanese and U.S. indices (markets) as an input vector and generates responses in the form of an output vector. The neural network training data covered the period from January 1, 1995, through December 31, 2000, resulting in approximately 2000 daily observations per series. The testing data covered the period from January 1, 2001, through December 31, 2001, resulting in approximately 250 daily observations per series. A filter-trading rule was used to verify the effectiveness of these predictions.;The empirical results confirmed that neural trading systems may produce superior information for the next trading day. If an investor were to trade on the opening position of a neural net-signaled trading day, this investor could significantly outperform the buy-and-hold strategy. However, if the investor were to trade on the closing position of a signaled trading day, superior profits could be lost. It is shown that signaled trading days present superior trading opportunities. If an investor were to use suggested signals to daytrade on opening and closing positions, he or she could significantly outperform other investment strategies. (Abstract shortened by UMI.). |
