Scholarlink

The paper investigates CITRONN (Crypto Image-based Trading On Neural Networks) for forecasting cryptocurrency market trends by converting OHLC time-series data into images using Gramian Angular Fields (GAF) and Markov Transition Fields (MTF). Testing on Binance data for 10 trading pairs over various horizons (7/30/90 days), CITRONN models yield high performance, particularly in short-term predictions (7 days) and generate Sharpe-ratio above 2.0, reaching 3.25 in some cases. Our research highlights CITRONN’s effectiveness in identifying complex patterns, improving prediction accuracy and trading strategies.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

CITRONN: A Convolutional Neural Network for Crypto Image-Based Trading

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Haiyun Zhu,
  • Beier Liu,
  • Mingjun Sun

摘要

The paper investigates CITRONN (Crypto Image-based Trading On Neural Networks) for forecasting cryptocurrency market trends by converting OHLC time-series data into images using Gramian Angular Fields (GAF) and Markov Transition Fields (MTF). Testing on Binance data for 10 trading pairs over various horizons (7/30/90 days), CITRONN models yield high performance, particularly in short-term predictions (7 days) and generate Sharpe-ratio above 2.0, reaching 3.25 in some cases. Our research highlights CITRONN’s effectiveness in identifying complex patterns, improving prediction accuracy and trading strategies.