An Ensemble Method of Deep Reinforcement Learning for Automated Cryptocurrency Trading

TL;DR

This paper introduces an ensemble approach using deep reinforcement learning to enhance cryptocurrency trading strategies, emphasizing robustness and adaptability in volatile markets through model selection, mixture policies, and periodic retraining.

Contribution

It presents a novel ensemble method with a mixture distribution policy and validation-based model selection to improve trading strategy robustness in dynamic cryptocurrency markets.

Findings

Enhanced out-of-sample performance over benchmarks

Robustness demonstrated across evolving market conditions

Effective handling of non-stationarity through periodic retraining

Abstract

We propose an ensemble method to improve the generalization performance of trading strategies trained by deep reinforcement learning algorithms in a highly stochastic environment of intraday cryptocurrency portfolio trading. We adopt a model selection method that evaluates on multiple validation periods, and propose a novel mixture distribution policy to effectively ensemble the selected models. We provide a distributional view of the out-of-sample performance on granular test periods to demonstrate the robustness of the strategies in evolving market conditions, and retrain the models periodically to address non-stationarity of financial data. Our proposed ensemble method improves the out-of-sample performance compared with the benchmarks of a deep reinforcement learning strategy and a passive investment strategy.