Pemilihan Parameter Crossover Moving Average Adaptif pada BTC/USDT Menggunakan Proximal Policy Optimization

Anandava Eka Buana Baskara (Universitas Teknologi Yogyakarta, Yogyakarta)
Joko Aryanto (Universitas Teknologi Yogyakarta, Yogyakarta)

TIN: TERAPAN INFORMATIKA NUSANTARA
Vol 6 No 12 (2026): May 2026

Cryptocurrency markets exhibit high volatility, making it challenging to determine the optimal combination of Moving Averages (MA Short and MA Long) for technical indicator-based trading strategies. This study aims to develop an adaptive crossover Moving Average strategy using the Proximal Policy Optimization (PPO) algorithm to evaluate and recommend the most effective MA combinations. Daily cryptocurrency price data from January 1, 2021, to May 15, 2026, comprising a total of 1960 candles, were obtained through an exchange platform API. The data were processed through preprocessing to form market states that include price, volume, and volatility indicators, which were then used as input for the PPO agent during training and strategy evaluation. Test results indicate that the MA 3/50 combination was most frequently selected by PPO based on average probability, while the MA 25/40 combination produced the best financial performance in terms of profit factor, net profit, and win rate. Visualizations of the equity curve, drawdown, and entry and exit points confirm the strategy’s ability to adaptively adjust decisions, capture market trends, and balance risk and profitability. These findings provide practical guidance for selecting adaptive crossover Moving Average parameters, enabling technical indicator-based trading strategies to navigate the complex and rapidly changing dynamics of cryptocurrency markets.