FORECASTING THE STOCK PRICES USING GENERATIVE ADVERSARIAL NETWORKS AND SENTIMENT ANALYSIS OF SOCIAL NETWORKS
DOI:
https://doi.org/10.31891/mdes/2022-6-4Keywords:
stock market, macroeconomic and social prerequisites, artificial neural networksAbstract
The study considers the problem of effective forecasting of stock prices in the context of global economic instability. Popular and simple forecasting methods are no longer effective in conditions of high market volatility, so scientists are testing new models, especially popular among which are neural networks. In recent years, Generative Adversarial Networks (GANs) have achieved promising results in solving many complex problems (e.g., creating realistic images and videos, image-to-image and text-to-image conversion), but the effectiveness of using this type of network for stock price forecasting is still a matter of debate. This type of models was previously used mainly to generate new photos, videos, or texts, but not time series, especially as volatile as stock prices. Over the past two years, researchers have started to test this type of networks for such tasks, but they still face high market volatility, which cannot be predicted using only historical data on the price and sale of shares. To help the network understand the bigger picture of the market we add different technical indicators to the training data, such as moving averages etc., which describe the development of stock price not only for the current day, but for the past week or more. But this still does not cancel the fact, that there is much more to the process of stock price formation. Over 1 day, one online post might be a turning point in the course of events, which may result in the market crash. Elon Musk tweets, coronavirus, start of russian full-scale invasion of Ukraine are the proof to that. Therefore, this paper will take into account another important external indicator, as the mood of stock market participants. The most effective method in this task is the analysis of the tone (sentiment analysis) of the text, in this article will be considered posts in the social network Twitter. Based on this, a model for forecasting stock prices will be created, which takes into account not only historical data and technical indicators, but also such external factors influencing the market as the mood of traders and brand reputation.
References
Romero R. Generative adversarial network for stock market price prediction. CD230: Deep Learning, Stanford University. 2018. P. 5.
Guoqiang Z. Stock market prediction based on generative adversarial network. 2019. URL: https://www.researchgate.net/publication/331002527_Stock_Market_Prediction_Based_on_Generative_Adversarial_Network/ (дата звернення: 29.10.2022).
Stock price prediction using generative adversarial networks / H. Lin et al. 2021. URL: https://thescipub.com/abstract/jcssp.2021.188.196/ (дата звернення: 29.10.2022).
Staffini A. Stock price forecasting by a deep convolutional generative adversarial network. 2022. URL: https://www.frontiersin.org/articles/10.3389/frai.2022.837596/full (дата звернення: 29.10.2022).
Mittal A., Goel A. Stock prediction using twitter sentiment analysis. CS 229. 2011. P. 5.
Sentiment analysis of twitter data for predicting stock market movements / V. Pagolu et al. International conference on Signal Processing, Communication, Power and Embedded System (SCOPES)-2016. 2016. URL: https://arxiv.org/pdf/1610.09225.pdf. (дата звернення: 29.10.2022).
Hutto C., Eric G. Vader: a parsimonious rule-based model for sentiment analysis of social media text. Advancement of Artificial Intelligence. 2014. P. 10.
