Stock price forecast based on improved Transformer

DOI

10.2991/978-94-6463-490-7_20How to use a DOI?

Keywords

stock; transformer; predicted; deep learning methods

Abstract

Nowadays, the stock trading market is expanding day by day due to economic globalization, and the stock trading data is increasing day by day. How to use more effective methods to select high quality stocks from numerous stock data has become an increasingly concerned issue for shareholders. Time series prediction has broad application prospects, which attracts more and more researchers to study it deeply. Moving average is a kind of technical index used to observe the trend of stock price change. It is one of the most widely used technical indexes and is often used in the task of predicting the trend of time series. But the original moving average indicator is obtained by calculating equal or preset weights assigned to the data of time series. Ignoring nuances of importance at different points in time; In addition, the same weight is used for various data of time series, ignoring the discrepancies in the intrinsic properties of different time series. The use of deep learning and machine learning methods to analyze and deal with stock trading data can effectively help shareholders to choose reasonable stocks for trading. Nowadays, many researchers have attempted to use deep learning methods of artificial-intelligence technology to deal with the question of stock price and trend prediction, though they have achieved good results, but they lack the stability of stock price prediction ability for different forecasting cycles. This paper adopts Transformer in deep learning technology as the basic network architecture to solve the above problems, to predict the closing price of stock prices, input stock prices at different time scales into the model, which can process more global time series feature information, and then use GRU technology to advance the features at different scales Row fusion enables the model to meet the predicted changes of data in different periods, this model is named MS-Transformer. In this study, the data of China Securities 50 stocks were selected as the experimental data. The results prove that under the effect factors of 5 basic trading indicators (opening price, trading volume, highest price, lowest price and closing price), MS-Transformer model can be proved to have good performance in stock price prediction.

Copyright

© 2024 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Hongwei Zhang
PY  - 2024
DA  - 2024/08/31
TI  - Stock price forecast based on improved Transformer
BT  - Proceedings of the 2024 3rd International Conference on Artificial Intelligence, Internet and Digital Economy (ICAID 2024)
PB  - Atlantis Press
SP  - 170
EP  - 182
SN  - 2589-4919
UR  - https://doi.org/10.2991/978-94-6463-490-7_20
DO  - 10.2991/978-94-6463-490-7_20
ID  - Zhang2024
ER  -