How clintjohnsn/rl-trader works
Project type and purpose
• This is a research‐oriented codebase supporting a Masters dissertation on stock trading using Group‐Agent Deep Reinforcement Learning (GARL).
• It implements a custom OpenAI Gym environment, several RL strategies (GARL, DQN, PPO, A2C) and an ARIMA baseline, plus data ingestion, backtesting and performance analysis.
(inferred) file structure
• Top‐level:
• Dissertation.pdf (write-up)
• README.md (project overview)
• LICENSE
• data/ (data ingestion & feature engineering)
• env/ (OpenAI Gym trading environment)
• agents/ (GARL implementation + single‐agent RL wrappers)
• baselines/ (ARIMA, buy-and-hold)
• backtest/ (simulation engine)
• analysis/ (performance metrics, plotting)
• config/ (hyperparameters, experiment specs)
README insights
• Core components: custom Gym environment, GARL strategy, two baseline strategies, data from Nasdaq-100 enriched with technical indicators, backtesting framework, performance analysis.
• External dependencies: OpenAI Gym, RL libraries (e.g., Stable Baselines), time‐series libraries (pandas, statsmodels).