Claim Missing Document
Check
Articles

Found 1 Documents
Search
Journal : Jurnal Mandiri IT

Reinforcement learning for bitcoin trading: A comparative study of PPO and DQN Prasetyo, Romadhan Edy; Sumanto, Sumanto; Chaidir, Indra; Supriyatna, Adi
Jurnal Mandiri IT Vol. 14 No. 2 (2025): Computer Science and Field
Publisher : Institute of Computer Science (IOCS)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35335/mandiri.v14i2.455

Abstract

Bitcoin’s high volatility demands automated strategies that adapt to changing market regimes while managing risk. This study compares Proximal Policy Optimization (PPO) and Deep Q-Network (DQN) for Bitcoin trading using hourly BTC/USDT data from 2019 to early 2025. The models are trained to generate buy and sell signals from technical indicators including the Relative Strength Index (RSI), MA20, volatility, Moving Average Convergence Divergence (MACD), volume trend, SMA200, and a weekly trend filter. All features are computed on hourly bars. The evaluation shows that PPO tends to trade more aggressively and delivers higher performance during bullish phases, though with greater risk in unstable markets. By contrast, DQN trades more selectively and maintains better stability in sideways or choppy conditions. These findings support the effectiveness of reinforcement learning for adaptive cryptocurrency trading and highlight complementary strengths between PPO and DQN across market regimes.