This repository contains the submission by Team Lean Large Men for the NTU Deep Learning Week 2025 competition, held from February 28, 2025, to March 3, 2025. Our project explores AI-driven trading strategies, leveraging deep reinforcement learning (DQN) and LSTM models to analyze and predict financial market movements.
Our work critically examines the challenges of algorithmic trading, acknowledging the complexities of financial markets where statistical relationships are constantly shifting. While outperforming institutional-grade quantitative firms remains unrealistic, our approach provides valuable insights into the application of deep learning in financial modeling.
- LSTM Model: Designed for price prediction, capturing both short-term and long-term dependencies in market data.
- Deep Q-Network (DQN) Reinforcement Learning Agent: An adaptive trading strategy that learns from market conditions to optimize decision-making.
To enhance model robustness, we incorporated macroeconomic and technical indicators such as:
- Moving Averages (50/200-day relationships)
- RSI (Relative Strength Index)
- MACD (Moving Average Convergence Divergence)
- Bollinger Bands
- Federal Interest Rates
- Global Liquidity Measures (M1/M2 Money Supply)
- US Dollar Strength Index
DL Week'25 Finance Track Report - Lean Large Men.pdf- Full project report detailing methodology, implementation, and results.LeanLargeMen_DLWeek2025.pptx- Presentation slides summarizing the project.LeanLargeMen_DLWeek2025.pdf- Presentation slides summarizing the project in PDF form.Deep RL approach.ipynb- Implementation of the Deep Q-Network (DQN) reinforcement learning model.LSTM approach.ipynb- Implementation of the LSTM-Transformer hybrid model for price prediction.
SPY_indicators.csv- Historical SPY price data with calculated technical indicators.fedIR.csv- Historical US Federal Reserve interest rates data.m1m2.xlsx- Historical data of M1 and M2 money supply measures.usdstrength.csv- US Dollar Index dataset tracking its strength against a basket of foreign currencies.final_df.csv- Merged dataset integrating all relevant macroeconomic and market indicators.
alphaVantage_marketSentiments.ipynb- API query implementation using AlphaVantage's market sentiment data.
We extend our gratitude to NTU Deep Learning Week 2025 organizers and mentors for providing a platform to explore AI in trading. This project was a valuable learning experience in financial modeling, deep learning, and reinforcement learning applications.
For any inquiries or discussions, feel free to open an issue or contact us through this repository!
📢 Happy Coding & Trading! 🚀