Stock price prediction using Generative Adversarial Networks
TL
TL;DR
This paper proposes a stock prediction model using Generative Adversarial Network (GAN) with Gated Recurrent Units (GRU) used as a generator that inputs historical stock price and generates future stock prices and Convolutional Neural Network (CNN) as a discriminator to discriminate between the real stockprice and generated stock price.
Abstract
Deep learning is an exciting topic. It has been utilized in many areas owing to its strong potential. For example, it has been widely used in the financial area which is vital to the society, such as high-frequency trading, portfolio optimization, fraud detection and risk management. Stock market prediction is one of the most popular and valuable areas in finance. In this paper, it proposes a stock prediction model using Generative Adversarial Network (GAN) with Gated Recurrent Units (GRU) used as a generator that inputs historical stock price and generates future stock price and Convolutional Neural Network (CNN) as a discriminator to discriminate between the real stock price and generated stock price. Different from the traditional methods, which limited the forecasting on one-step-ahead only, by contrast, using the deep learning algorithm is possible to conduct the multi-step ahead prediction more accurately. In this study, it chose the Apple Inc. stock closing price as the target price, with features such as S&P 500 index, NASDAQ Composite index, U.S. Dollar index, etc. In addition, FinBert has been utilized to generate a news sentiment index for Apple Inc. as an additional predicting feature. Finally, this paper compares the proposed GAN model results with the baseline model.
References8 items
1
Stock Market Prediction on High-Frequency Data Using Generative Adversarial Nets
2
Rainbow: Combining Improvements in Deep Reinforcement Learning
3
Proximal Policy Optimization Algorithms
4
Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation
5