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

ArXiv ID: 2410.19291 “View on arXiv”

Authors: Unknown

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%.

Keywords: Convolutional Neural Network (CNN), Image-based Prediction, Stock Trend, Time Series, Multi-scale Fusion, Equities

Complexity vs Empirical Score

  • Math Complexity: 4.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Street Traders
  • Why: The paper’s mathematics is moderate, relying primarily on standard CNN architectures and time series decomposition rather than advanced theoretical derivations. However, it demonstrates high empirical rigor through extensive experiments on 4,454 A-share stocks, reporting specific metrics like 61.15% positive predictive value and 165.09% total profit, indicating backtest-ready implementation and substantial data/effort.
  flowchart TD
    A["Research Goal<br>Improve Stock Price Prediction<br>Accuracy for China A-Share Market"] --> B["Data Preparation<br>4,454 A-Share Stocks<br>Open/High/Low/Close/Time Images"]
    B --> C["Methodology: SMSFR-CNN<br>Sequence-based Multi-scale Fusion<br>Regression CNN"]
    C --> D["Computational Process<br>1. CNN Feature Extraction<br>2. Multi-scale Fusion<br>3. Regression Prediction"]
    D --> E["Key Findings<br>PPV: 61.15% | NPV: 63.37%<br>Total Profit: 165.09%<br>Reduced Overfitting"]