StockGAN: robust stock price prediction using GAN algorithm

Abstract

Stock market predictions help investors benefit in the financial markets. Various papers have proposed different techniques in stock market forecasting, but no model can provide accurate predictions. In this study, we show how to accurately anticipate stock prices using a prediction model based on the Generative Adversarial Networks (GAN) method. We collect the dataset, preprocess it, extract features, evaluate the model, and then deploy the GAN method to develop a stock price prediction model. The GAN comprises two parts: a generator and a discriminator that are both trained using adversarial learning processes. In this study, we utilize features including date, open, high, low, close, and volume to train our model. The results of the experiments gain good accuracy and a low error rate, so it can be a promising solution for dealing with accurate and dynamic stock prices. Moreover, the proposed model can achieve the results obtained are a metric score r2 with real predictions = 0.811166 and synthetic predictions = 0.674971. The MAE function produces real predictions = 0.020665, and synthetic predictions = 0.042406. The MRLE gains real = 0.001087 and synthetic = 0.002479.