TradeBotRL - Smarter Trading Decisions, Simplified 📈🤖

TradeBot is a reinforcement learning-powered trading assistant designed to help users make informed and optimized trading decisions. By leveraging sentiment analysis and market data, TradeBot dynamically adjusts buy and sell thresholds to maximize portfolio performance. This tool is ideal for traders seeking to integrate artificial intelligence into their strategies, offering insights based on real-time sentiment and price trends.

How It is Built 👷

Tech Stack 💻

• Python for backend development and reinforcement learning implementation.
• Stable-Baselines3 PPO for reinforcement learning model training.
• Gymnasium for creating a custom trading environment.
• Pandas & NumPy for data handling and preprocessing.
• Sentiment Analysis Model: Hugging Face's twitter-roberta-base-sentiment-latest for analyzing sentiment scores from social media data.

How It Works 🧑‍🍳

1. Data Collection:
   • Scrape Reddit posts using reddit_scrape.py or Mastodon posts with mastodon_scraping.py. Specify the start/end dates and subreddit or hashtag.
2. Sentiment Analysis:
   • Analyze collected posts with sentiment_analysis.py, which uses a pre-trained sentiment model to generate sentiment scores.
3. Trading Strategy Implementation:
   • The custom Gymnasium environment (TradingEnv) simulates trading decisions based on sentiment scores and market data (e.g., closing prices).
4. Reinforcement Learning:
   • Train a Proximal Policy Optimization (PPO) model to optimize buy/sell thresholds dynamically.
5. Backtesting Framework:
   • Use backtesting_with_sentiment.py to evaluate the basic performance, generating metrics and transaction logs.
   • use testingRL.py to evaluate automatic trade bot, generating action steps and final value

Usage 🍳

1. Prepare your dataset by running the scraping scripts (reddit_scrape.py or mastodon_scraping.py) and analyzing them with sentiment_analysis.py.
2. Train the PPO model using market data in TradingEnv.
3. Deploy the trained model to simulate trading or use it in real-world scenarios.

How to Run the Program 💻

Environment Setup

1. Clone the repository and navigate to the project folder.
2. Install required libraries:

> pip install -r requirements.txt

Data collection

1. Use reddit_scrape.py and mastodon_scraping.py to collect post data from reddit and mastodon
2. Use sentiment_analysis.py to categorize the sentiment of each post content
3. Use data_preprocessing.py and necessary script to clean the data

Train and test RL agent

1. Train the model by tesla stock data using the RLbot.py file, altering the timestep to train to your needs
2. Test the model by nvidia stock data using the testingRL.py file

License

TradeBotRL is licensed under MIT License. All development is currently maintain by Hoang Quy Nguyen.

Disclaimer: Use this bot at your own risk