Decomposition-Enhanced Network for Financial Time Series Forecasting

This repository contains the implementation of the DENet (Decomposition-Enhanced Network) model, a novel forecasting architecture designed to drive the R-breaker strategy for enhanced profitability in futures trading. The project evaluates the model's performance across five different futures contracts using both daily and minute-level data. Traditional trading strategies often rely on static price levels. Our approach introduces DENet, a prediction model that dynamicizes the R-breaker strategy. By forecasting price movements, DENet allows the strategy to preemptively adjust entry and exit points, significantly improving risk-adjusted returns. Key Contributions: DENet Architecture: A specialized deep learning model for time-series forecasting in futures markets. Hybrid Strategy: Integration of DENet predictions with the classic R-breaker logic . Multi-Granularity Analysis: Comprehensive testing on both daily and minute-level frequencies. Significance Testing: Statistical validation of the trading results to ensure robustness.