Stock Market Price Prediction Based on TGAN

The use of deep learning methods offers several advantages for processing big data, making it applicable to various real-life fields such as finance, industry, medicine, and sales. This paper proposes a novel model called Temporal Convolutional Networks-Generative Adversarial Nets (TGAN). The generator component of the model employs Temporal Convolutional Networks (TCN), while the discriminator component is made up of Convolutional Neural Networks (CNN). We conduct an ablation experiment using TGAN and compare its performance with other models such as the Differential Autoregressive Integrated Moving Average Model (ARIMA), Long-Short Term Memory (LSTM), and Gated Recurrent Unit (GRU) models. The experiment involves analyzing the trend of Apple's stock closing price from July 2010 to June 2020, using the Fourier transform method. The experimental results show that the TGAN model outperforms the other models in terms of prediction accuracy, with Root Mean Square Error (RMSE) values of 1.25 and 1.63 for single-step and multi-step prediction of stock closing prices, respectively. We further conducted experiments on three different types of stock data to verify the model's performance and generalization, and the results demonstrate that TGAN is practical and effective.