Integrating Fibonacci Retracement To Improve Accuracy of Time Series Prediction of Gold Prices

Bagus Priambodo (Universitas Mercu Buana)
Ruci Meiyanti (Universitas Mercu Buana)
Samidi Samidi (Universitas Budi Luhur)
Gushelmi Gushelmi (Universitas Putra Indonesia YPTK)
Rabiah Abdul Kadir (Universiti Kebangsaan Malaysia)
Azlina Ahmad (Universiti Kebangsaan Malaysia)

Journal of Applied Engineering and Technological Science (JAETS)
Vol. 6 No. 2 (2025): Journal of Applied Engineering and Technological Science (JAETS)

The prediction of gold prices is crucial for investors and policymakers due to its significant impact on global financial markets. Machine learning and deep learning have been used for predicting gold prices on time series data. This study employs MLR, SVM and CNN LSTM with Fibonacci retracement levels to forecast gold prices based on time series data. The experiment results demonstrate that combining Fibonacci retracement with model prediction significantly enhances predictive performance compared to prediction without Fibonacci. The use of Fibonacci levels has resulted in a higher R² score and lower RMSE score showing that Fibonacci levels influence the accuracy of gold price predictions and strengthen the overall reliability of gold price forecasts. The findings underscore the potential of combining machine learning models with technical analysis tools in financial forecasting. Integrating the Fibonacci retracement level offers valuable insights for market participants, enabling more informed investment decisions and effective risk management strategies.