Awesome AI Agents › Financial & Trading Systems

pskrunner14/trading-bot

⭐ 1180 Jupyter Notebook repository created 2018-08-13

This project is a Stock Trading Bot that leverages Deep Reinforcement Learning, specifically Deep Q-Learning, to make trading decisions. The bot is designed to learn optimal trading policies by interacting with stock market data, using a model-free reinforcement learning approach. The implementation closely follows the algorithmic principles discussed in foundational research papers, making it a valuable resource for those interested in understanding and applying Deep Q-Learning in financial trading contexts. The bot operates by observing the current state of the stock market, represented as an n-day window of stock price changes normalized through a sigmoid function. Based on this state, the agent decides whether to buy, sell, or hold a stock. The agent then receives a reward based on the change in portfolio value, which it uses to update its decision-making model. This iterative process allows the bot to improve its trading strategy over time. Several enhancements to the basic Q-Learning algorithm have been implemented, including Vanilla DQN, DQN with fixed target distribution, and Double DQN. However, some advanced features like Prioritized Experience Replay and Dueling Network Architectures are noted as planned but not yet implemented. The bot has been trained on historical stock data from Google (GOOG) spanning 2010 to 2017 and tested on data from 2019, achieving notable profits. The project includes visualization tools to analyze the bot's performance and provides scripts for training and evaluation, making it accessible for users to experiment with their own data. The project emphasizes simplicity and educational value, making it suitable for those learning about reinforcement learning applications in finance. It also acknowledges the limitations of the current approach, such as the restriction to trading one stock at a time and the preference for CPU-based training due to the sequential nature of the learning process. Overall, this trading bot project offers a practical and educational implementation of Deep Q-Learning for stock trading, supported by historical data, visualization tools, and clear documentation.

https://github.com/pskrunner14/trading-bot

ai-agentsbuy-sell-hold-actionsdeep-q-learningdeep-reinforcement-learningdouble-dqnfinancial-tradinggoogle-stockhistorical-stock-datamachine-learningmodel-free-reinforcement-learningneural-networksportfolio-managementq-learningreinforcement-learningreward-signalstate-representationstock-marketstock-price-predictionstock-tradingstock-trading-bottrading-agenttrading-algorithmstrading-bottrading-strategytraining-and-evaluationvanilla-dqnvisualization