Skip to content
Yinliang Chen
All projects

BitHunter: Leveraging LLMs for Cryptocurrency Trading Insights

image
May 2, 2025

Overview

BitHunter investigates using Large Language Models (LLMs) to analyze textual data (news, social media, forums) for cryptocurrency trading signals (Buy/Sell/Hold). This MVP converts LLM-driven insights into trading decisions, tested with simulated/historical data in a backtesting framework, designed for safe testing and future enhancements like live data.

Key Features

  • Data Ingestion: Fetches historical crypto price data and news (simulated/historical in MVP).
  • LLM Sentiment Analysis: Processes text for market sentiment (keyword heuristics in MVP, planned full LLM integration e.g., GPT API).
  • Signal Generation: Generates trading signals from sentiment scores and other indicators.
  • Backtesting Engine: Simulates signal execution against historical price data to evaluate strategy performance and risk.
  • Dashboard: Flask-based web interface to visualize data, signals, and backtesting results.

Technologies Used

  • AI/ML: Conceptual LLM integration (e.g., GPT API planned), keyword-based heuristics for MVP.
  • Backend: Python for core modules (Data Ingestion, Sentiment Analysis, Signal Generation, Backtesting).
  • Web Dashboard: Flask.
  • Data: Utilized historical and simulated datasets for development and testing.

Challenges and Learnings

  • Modular Design: Realized the critical importance of modular architecture in complex systems for maintainability and scalability.
  • Simulated Data Benefits: Recognized the significant advantages of starting with simulated data for iterative development and validating core logic before live system integration.
  • Sentiment to Signal Translation: Understood the complexities of converting qualitative sentiment analysis into quantitative and reliable trading signals.

Outcome

The BitHunter MVP successfully demonstrated the feasibility of translating text analysis (simulated LLM) into trading signals and evaluating them via backtesting. This establishes a strong foundation for future developments, including live data/LLM API integration, algorithm refinement with more technical indicators, advanced risk management, enhanced interactive dashboard features, and potential exploration of reinforcement learning for optimizing trading decisions.