Stock Market Prediction on High-Frequency Data Using Generative Adversarial Nets

Abstract

Stock price prediction is an important issue in the financial world, as it contributes to the development of effective strategies for stock exchange transactions. In this paper, we propose a generic framework employing Long Short-Term Memory (LSTM) and convolutional neural network (CNN) for adversarial training to forecast high-frequency stock market. This model takes the publicly available index provided by trading software as input to avoid complex financial theory research and difficult technical analysis, which provides the convenience for the ordinary trader of nonfinancial specialty. Our study simulates the trading mode of the actual trader and uses the method of rolling partition training set and testing set to analyze the effect of the model update cycle on the prediction performance. Extensive experiments show that our proposed approach can effectively improve stock price direction prediction accuracy and reduce forecast error.

RNNLBL : A Recurrent Neural Network and Log Bilinear based Efficient Stock Forecasting Model

“…This section present performance evaluation of proposed RNNLBL stock market forecasting model over existing stock market forecasting model [17]. The existing model is designed using convolution neural network.…”

“…The performance of proposed RNNLBL model and existing stock forecasting model is evaluated in terms of (Root Mean Squared Relative Error) RMSRE and Direction Prediction Accuracy (DPA). The experiment is conducted using china stock exchange (CSE) data used in [17] which composed of extreme volatility and accidental event extremely impacting stock price. The CSE data ranges from January 1, 2016 to December 31 st , 2016 that composed of total 244 trading days and each day is composed of 242-minute sessions with respect to 59048 time periods.…”

“…The process is repeated till the entire stock data is forecasted (i.e., can be used to updated the learning model and depicts the size of respective training stock dataset). More details of dataset can be obtained from [17]. Experiments are conducted for different stock considering time , a forecasting is done for next interval of time using proposed learning methods.…”

“…Meanwhile any inappropriate setting in parameter leads to the disaster performance of model; hence, parameter setting is the real scenario optimization problem. Moreover, to address this issue, [17] proposed the direction forecasting loss and forecasting error loss and it is observed that GA (Generative Adversarial) training can be applied to integrate the loss parameter for obtaining the sustaining outcomes. Moreover, this prediction architecture is named as GAN for forecast error loss and direction prediction loss aka GAN-FD.…”

“…Moreover, stock price is predicted mainly to determine the company's future value or the other financial assets that can be marketed on the exchange [3] [7]. Meanwhile it is also observed that the stock market is heavily characterized through the various properties such as high frequency MC (Multi-polynomial Component), nonlinearities and discontinuities since fluctuations occurs due to the various scenario such as political changes, Country economic conditions and the expectation of traders [4] [17], hence it is highly improbable to predict the stock values. Moreover the investors buy the stocks which deals with the infrastructure projects, construction firms, contractors hiring, paperwork handling.…”