Pengembangan dan Optimasi Kalman Filter Adaptif Multi-Timeframe dalam Kerangka State-Space untuk Prediksi Arah Harga Valuta Asing: Development and Optimization of Multi-Timeframe Adaptive Kalman Filter in State-Space Framework for Foreign Exchange Price Direction Prediction

Baradja, Abdillah
Sukoco, Sukoco
Mukti, Bayu

Jurnal Pendidikan Sains dan Komputer
Vol. 6 No. 01 (2026): Artikel Riset, February 2026

The foreign exchange market is non-stationary, noisy, and highly sensitive to changes in information, making short-term price direction prediction a significant challenge. Although the Kalman Filter has been widely used within a state-space framework for trend and dynamic parameter estimation, most previous research has focused on statistical accuracy without directly linking it to actual trading performance. This study aims to develop and evaluate an adaptive Kalman Filter-based forex price direction prediction model and empirically test its implications for the performance of algorithmic trading systems. The study uses a quantitative experimental approach, analyzing historical EUR/USD data across the H1, H2, H3, and H4 timeframes over 1 year. The state-space-based Kalman Filter model is implemented in a MetaTrader 5 Expert Advisor and compared with the Moving Average Crossover strategy and the classic Kalman Filter. Evaluations are conducted at the prediction and trading levels using Total Net Profit, Profit Factor, Sharpe Ratio, and Drawdown. The proposed model demonstrates consistently positive performance across all timeframes, with a Profit Factor of 1.39–2.13 and a Sharpe Ratio of up to 5.70, significantly outperforming the technical baseline and the classic Kalman Filter, which exhibit unstable results. The integration of the Kalman Filter approach into an algorithmic trading system improved signal quality and risk efficiency during testing.