Generative adversarial network (GAN) and enhanced root mean square error (ERMSE): deep learning for stock price movement prediction

The prediction of stock price movement direction is significant in financial circles and academic. Stock contains complex, incomplete, fuzzy information which makes it an extremely difficult task to predict its development trend. Predicting analysing data a nonlinear, time-dependent problem. With rapid machine learning deep learning, this can be performed more effectively by purposely designed network. This paper aims improve accuracy minimizing forecasting error loss through architecture using Generative Adversarial Networks. It was proposed generic model consisting Phase-space Reconstruction (PSR) method for reconstructing series Network (GAN) combination two neural networks are Long Short-Term Memory (LSTM) as Convolutional Neural (CNN) Discriminative adversarial training forecast the market. LSTM will generate new instances based on historical basic indicators then CNN estimate whether predicted or real. found that has well enhanced root mean square LSTM, 4.35% accurate predicting reduced processing time RMSE 78 secs 0.029, respectively. study provides better result index. seems system concentrates improving accuracy,