本研究利用倒傳遞演算法為架構的前饋式類神經網路，以香港掛牌ETF 及香港跨境台灣掛牌ETF 為研究標的，探討其輸入變數香港恒生指數前一日收盤價是否有「價格發現」功能，以提供給國內外投資人作為決策輔助系統以獲取較佳投資績效。因為香港股票市場收盤時間為16：00，而台灣股票市場收盤時間13：30，造成香港收盤時間較晚，所以本研究依據相同的輸入變數香港恒生指數前一日收盤價，分別與輸出變數香港掛牌ETF 當日開盤價、香港跨境台灣掛牌ETF 當日開盤價兩者互相比較驗證，是否仍具有相同股價預期能力。實證結果顯示，香港恒生指數前一日收盤價對香港掛牌ETF 當日開盤價預期能力較對香港跨境台灣掛牌ETF 當日開盤價預期能力來得好，這也表示香港恒生指數前一日收盤價對香港掛牌ETF 當日開盤價相較於對香港跨境台灣掛牌ETF 當日開盤價有價格發現功能。

The study uses the back-propagation algorithm for the structure of feed-forward neural network, The purpose of this research focuses on the Hong Kong ETF listed in Hong Kong and the Hong Kong ETF cross-border listed in Taiwan, and explore its input (forecast) variables, whether there is [price discovery] function, to provide for domestic and foreign investors as a decision-support systems to obtain better investment performance. Because Hong Kong stock market’s closing time is 16:00, while Taiwan stock market’s closing time is 13:30, resulting in Hong Kong to close later. Therefore this study was based on the same input (forecast) variables, respectively, and output (target) variables Hong Kong ETF listed in Hong Kong and the Hong Kong's ETF cross-listed in Taiwan to compare the two verifications, the expected stock price still has the same capacity.
The empirical results show that neural network is expected to verify the ability of Hong Kong ETF listed in Hong Kong stock price forecast capacity was better than in Hong Kong ETF cross-border Listed in Taiwan stock price forecast capacity.