| Summary: | 碩士 === 輔仁大學 === 金融與國際企業學系金融碩士班 === 106 === This study uses the intraday data information to forecasts the trend of stock prices within one day by using Long Short-Term Memory(LSTM) neural network model, and refers to the previous literature in Parlour(1998), Bacidore et al(2003) Cao and Wang(2003) ,the data of limit order book are adopted to as input parameters. The data period is from December 1, 2017 to December 31, 2017. Every data in limit order book will be collected in each transaction. In this study we will discussed the degree of influence of the limit order book information and the prediction ability of the LSTM neural network. The underlying was selected from Taiwan Stock Exchange electronic stock index constituent stocks, which price has the highest growth rate in 2017, Wang Hong (Stock code 2337) and TSMC (Stock code 2330). We will use them to test the predictive power of LSTM in high volatility and low volatility value stocks. The empirical study shows that, no matter how fluctuated the stock amplitude was, the LSTM neural network is not affected by the size of stock fluctuations and can accurately predict the price movements. Also enabling investors to make buying and selling decisions based on the future price trend shown by the model.
|
|---|
