Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DFT: A Dual-branch Framework of Fluctuation and Trend for Stock Price Prediction

Chengqi Dong, Zhiyuan Cao, S Kevin Zhou, Jia Liu

TL;DR

A Dual-branch Framework of Fluctuation and Trend (DFT), which decomposes stocks into trend and fluctuation components and effectively extracts short-term fluctuations and trend information from stocks while explicitly modeling temporal variations and causal correlations.

Abstract

Stock price prediction is of significant importance in quantitative investment. Existing approaches encounter two primary issues: First, they often overlook the crucial role of capturing short-term stock fluctuations for predicting high-volatility returns. Second, mainstream methods, relying on graphs or attention mechanisms, inadequately explore the temporal relationships among stocks, often blurring distinctions in their characteristics over time and the causal relationships before and after. However, the high volatility of stocks and the intricate market correlations are crucial to accurately predicting stock prices. To address these challenges, we propose a Dual-branch Framework of Fluctuation and Trend (DFT), which decomposes stocks into trend and fluctuation components. By employing a carefully design decomposition module, DFT effectively extracts short-term fluctuations and trend information from stocks while explicitly modeling temporal variations and causal correlations. Our extensive experiments demonstrate that DFT outperforms existing methods across multiple metrics, including a 300% improvement in ranking metrics and a 400% improvement in portfolio-based indicators. Through detailed experiments, we provide valuable insights into different roles of trends and fluctuations in stock price prediction.

DFT: A Dual-branch Framework of Fluctuation and Trend for Stock Price Prediction

TL;DR

A Dual-branch Framework of Fluctuation and Trend (DFT), which decomposes stocks into trend and fluctuation components and effectively extracts short-term fluctuations and trend information from stocks while explicitly modeling temporal variations and causal correlations.

Abstract

Stock price prediction is of significant importance in quantitative investment. Existing approaches encounter two primary issues: First, they often overlook the crucial role of capturing short-term stock fluctuations for predicting high-volatility returns. Second, mainstream methods, relying on graphs or attention mechanisms, inadequately explore the temporal relationships among stocks, often blurring distinctions in their characteristics over time and the causal relationships before and after. However, the high volatility of stocks and the intricate market correlations are crucial to accurately predicting stock prices. To address these challenges, we propose a Dual-branch Framework of Fluctuation and Trend (DFT), which decomposes stocks into trend and fluctuation components. By employing a carefully design decomposition module, DFT effectively extracts short-term fluctuations and trend information from stocks while explicitly modeling temporal variations and causal correlations. Our extensive experiments demonstrate that DFT outperforms existing methods across multiple metrics, including a 300% improvement in ranking metrics and a 400% improvement in portfolio-based indicators. Through detailed experiments, we provide valuable insights into different roles of trends and fluctuations in stock price prediction.

Paper Structure

This paper contains 31 sections, 8 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Stock Trend Prediction.
  4. Stock Correlation Prediction.
  5. Methodology
  6. Overview
  7. Problem definition
  8. Stock decomposition
  9. Input embedding layer.
  10. Information decomposition module.
  11. Correlation modeling
  12. Fluctuation branch.
  13. Trend branch.
  14. Output prediction
  15. Experiments
  16. ...and 16 more sections

Figures (5)

  • Figure 1: Overall performance comparsion on the CSI800 and S&P500 stock datasets.
  • Figure 2: Our model DFT structure, a framework for independently modeling the correlation of time and stock dimensions in fluctuation and trend information.
  • Figure 3: The RWKV module structure.
  • Figure 4: Cumulative portfolio returns on the CSI800 and S&P500 test sets. Benchmark represents the CSI300 and S&P500 market indices, respectively.
  • Figure 5: The results of hyperparameter sensitivity. The x-axis in the plots represents Avgpool kernel size, convolutional kernel size, RWKV heads, lookback window length, and predict interval, respectively. The results of (a), (b) and (c) are obtained in the CSI800 dataset.