Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Stock Price Prediction Approach Based on Time Series Decomposition and Multi-Scale CNN using OHLCT Images

Zhiyuan Pei, Jianqi Yan, Jin Yan, Bailing Yang, Ziyuan Li, Lin Zhang, Xin Liu, Yang Zhang

TL;DR

This paper proposes a novel method, named Sequence-based Multi-scale Fusion Regression Convolutional Neural Network (SMSFR-CNN), for predicting stock price movements in the China A-share market, utilizing CNN to learn sequential features and combining them with image features to improve the accuracy of stock trend prediction.

Abstract

Recently, deep learning in stock prediction has become an important branch. Image-based methods show potential by capturing complex visual patterns and spatial correlations, offering advantages in interpretability over time series models. However, image-based approaches are more prone to overfitting, hindering robust predictive performance. To improve accuracy, this paper proposes a novel method, named Sequence-based Multi-scale Fusion Regression Convolutional Neural Network (SMSFR-CNN), for predicting stock price movements in the China A-share market. By utilizing CNN to learn sequential features and combining them with image features, we improve the accuracy of stock trend prediction on the A-share market stock dataset. This approach reduces the search space for image features, stabilizes, and accelerates the training process. Extensive comparative experiments on 4,454 A-share stocks show that the model achieves a 61.15% positive predictive value and a 63.37% negative predictive value for the next 5 days, resulting in a total profit of 165.09%.

A Stock Price Prediction Approach Based on Time Series Decomposition and Multi-Scale CNN using OHLCT Images

TL;DR

This paper proposes a novel method, named Sequence-based Multi-scale Fusion Regression Convolutional Neural Network (SMSFR-CNN), for predicting stock price movements in the China A-share market, utilizing CNN to learn sequential features and combining them with image features to improve the accuracy of stock trend prediction.

Abstract

Recently, deep learning in stock prediction has become an important branch. Image-based methods show potential by capturing complex visual patterns and spatial correlations, offering advantages in interpretability over time series models. However, image-based approaches are more prone to overfitting, hindering robust predictive performance. To improve accuracy, this paper proposes a novel method, named Sequence-based Multi-scale Fusion Regression Convolutional Neural Network (SMSFR-CNN), for predicting stock price movements in the China A-share market. By utilizing CNN to learn sequential features and combining them with image features, we improve the accuracy of stock trend prediction on the A-share market stock dataset. This approach reduces the search space for image features, stabilizes, and accelerates the training process. Extensive comparative experiments on 4,454 A-share stocks show that the model achieves a 61.15% positive predictive value and a 63.37% negative predictive value for the next 5 days, resulting in a total profit of 165.09%.

Paper Structure

This paper contains 17 sections, 9 equations, 5 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Materials and Methods
  4. Data Modalities
  5. Proposed Methods
  6. Multi-scale Cascading Feature
  7. Multi-Scale Resolution (MSR) CNN
  8. Sequence Multi-Scale Fusion Regression (SMSFR) CNN
  9. Experiments
  10. China A-shares Stock Data
  11. Classification Experiments
  12. Experimental Setup
  13. Comparison Between Time Series and Image Models on Original Stock Data
  14. Multi-Scale Feature Decomposition on Stock Data
  15. Profitability Backtest Experiments
  16. ...and 2 more sections

Figures (5)

  • Figure 1: Illustrations of the OHLCV and TS-OHLCT chart in (a) and (b), respectively.
  • Figure 2: Illustration of the multi-scale decomposition process of a 60-day feature map. Each sub-feature map transition from high-resolution local recent data to low-resolution long-term global trends as $i$ increases.
  • Figure 3: Our proposed network architecture of MSR-CNN.
  • Figure 4: Our proposed network architecture of SMSFR-CNN.
  • Figure 5: Profit comparison among time series baselines, reproduced OHLCV-based CNN, our proposed MSR and SNSFR models, and three market indexes on the test set.