A Hybrid Machine Learning Framework for Systematic Trading in Cryptocurrency and FX Markets

Fuente: Zenodo
Salvato in:
Dettagli Bibliografici
Autore principale: Izzuddin, Muchammad Fikri
Natura: Recurso digital
Lingua:inglese
Pubblicazione: Zenodo 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866902126083440640
author Izzuddin, Muchammad Fikri
author_facet Izzuddin, Muchammad Fikri
contents <p>This paper introduces a <strong>hybrid machine learning framework for algorithmic trading</strong> in highly volatile <strong>cryptocurrency</strong> and <strong>foreign exchange (FX) markets</strong>. The framework integrates <strong>high-frequency market data (OHLCV)</strong>, <strong>order book microstructure signals</strong>, and <strong>sentiment analysis powered by Large Language Models (LLMs)</strong> to generate systematic trading strategies.</p> <p>The approach leverages a multi-model architecture: <strong>LightGBM</strong> for efficient feature learning, <strong>LSTM and Transformer neural networks</strong> for capturing temporal dependencies, and a <strong>regime-switching mechanism</strong> to adapt across different market conditions. Final trading signals are combined through a <strong>stacking ensemble method</strong> to maximize robustness and predictive accuracy.</p> <p>To ensure practical application, the framework incorporates <strong>walk-forward validation</strong>, <strong>realistic transaction cost and slippage modeling</strong>, and a <strong>risk management overlay</strong> based on ATR stops, volatility targeting, and maximum drawdown controls. Backtests using data from <strong>2019–2024</strong> show significant improvements in <strong>Sharpe ratio, drawdown reduction, and profit factor</strong>, outperforming individual models and benchmark strategies.</p> <p>Results highlight the <strong>importance of LLM-driven sentiment features</strong>, which provide measurable improvements in predictive power and trading performance. This work contributes to the growing literature on <strong>machine learning in finance, quantitative trading, and AI-driven systematic strategies</strong>, offering a <strong>scalable and adaptable solution for crypto trading, forex trading, and algorithmic portfolio management</strong>.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_16966978
institution Zenodo
language eng
publishDate 2025
publisher Zenodo
record_format zenodo
spellingShingle A Hybrid Machine Learning Framework for Systematic Trading in Cryptocurrency and FX Markets
Izzuddin, Muchammad Fikri
Machine learning
Machine Learning
Algorithmic Trading
Cryptocurrency
Forex Trading
Quantitative Trading
Hybrid Models
Deep Learning
LLM Sentiment
Systematic Trading
<p>This paper introduces a <strong>hybrid machine learning framework for algorithmic trading</strong> in highly volatile <strong>cryptocurrency</strong> and <strong>foreign exchange (FX) markets</strong>. The framework integrates <strong>high-frequency market data (OHLCV)</strong>, <strong>order book microstructure signals</strong>, and <strong>sentiment analysis powered by Large Language Models (LLMs)</strong> to generate systematic trading strategies.</p> <p>The approach leverages a multi-model architecture: <strong>LightGBM</strong> for efficient feature learning, <strong>LSTM and Transformer neural networks</strong> for capturing temporal dependencies, and a <strong>regime-switching mechanism</strong> to adapt across different market conditions. Final trading signals are combined through a <strong>stacking ensemble method</strong> to maximize robustness and predictive accuracy.</p> <p>To ensure practical application, the framework incorporates <strong>walk-forward validation</strong>, <strong>realistic transaction cost and slippage modeling</strong>, and a <strong>risk management overlay</strong> based on ATR stops, volatility targeting, and maximum drawdown controls. Backtests using data from <strong>2019–2024</strong> show significant improvements in <strong>Sharpe ratio, drawdown reduction, and profit factor</strong>, outperforming individual models and benchmark strategies.</p> <p>Results highlight the <strong>importance of LLM-driven sentiment features</strong>, which provide measurable improvements in predictive power and trading performance. This work contributes to the growing literature on <strong>machine learning in finance, quantitative trading, and AI-driven systematic strategies</strong>, offering a <strong>scalable and adaptable solution for crypto trading, forex trading, and algorithmic portfolio management</strong>.</p>
title A Hybrid Machine Learning Framework for Systematic Trading in Cryptocurrency and FX Markets
topic Machine learning
Machine Learning
Algorithmic Trading
Cryptocurrency
Forex Trading
Quantitative Trading
Hybrid Models
Deep Learning
LLM Sentiment
Systematic Trading
url https://doi.org/10.5281/zenodo.16966978