QF-with-Reinforce-Learning

QF-with-Reinforce-Learning Reinforce Quantitive Finance automatic trading program. This project uses A2C (Advantage Actor-Critic) models trained on real cryptocurrency chart data to execute trades via Binance API and send notifications through Telegram.

---

Project Overview

• trade.py – Executes real trades based on trained models (Binance API + Telegram integration).
• train.py – Implements training algorithms (Reinforcement Learning, A2C).
• Uses 24/7 cryptocurrency chart data for training and evaluation.
• Includes both backtesting results and real-world performance tests.

Features

• Automated trading with reinforcement learning (A2C).
• Binance API integration for live trading.
• Telegram bot notifications for trade signals and execution logs.
• Backtesting and real-world evaluation to validate performance.

How to Install & Use

1. Environment Setup

Use Python 3.9 version.

Configuration

Update trade.py -> BASE_DIR to the current project path. Add your Binance API keys and Telegram bot token/chat ID inside the users/ folder.

Install Required Packages

# Required Packages
pip install tf-nightly
pip install numpy==1.26.0
pip install python-telegram-bot==13.14
pip install python-binance

# Start trading (background execution)
nohup python3 trade.py &

# Train models
python3 train.py train

# Test models
python3 train.py test

Traning & Trading Results

• Traning Result: https://softcon.ajou.ac.kr/works/works_prev.asp?uid=1730&wTerm=2024-1
• Real trading Result: 20% loss during six monts.