PERAMALAN HARGA EMAS INDONESIA MENGGUNAKAN XGBOOST DENGAN PENDEKATAN MULTI-VARIABEL MAKROEKONOMI DAN INDIKATOR TEKNIKAL

Akbar, Nasrul Fadhila (2026) PERAMALAN HARGA EMAS INDONESIA MENGGUNAKAN XGBOOST DENGAN PENDEKATAN MULTI-VARIABEL MAKROEKONOMI DAN INDIKATOR TEKNIKAL. Undergraduate thesis, UPN Veteran Jawa Timyr.

	Text 22082010182-cover.pdf Download (833kB)
	Text 22082010182-bab 1.pdf Download (423kB)
	Text 22082010182-bab 2.pdf Restricted to Repository staff only until 20 July 2029. Download (515kB)
	Text 22082010182-bab 3.pdf Restricted to Repository staff only until 20 July 2029. Download (411kB)
	Text 22082010182-bab 4.pdf Restricted to Repository staff only until 20 July 2029. Download (1MB)
	Text 22082010182-bab 5.pdf Download (246kB)
	Text 22082010182-daftarpustaka.pdf Download (208kB)

Abstract

Gold is one of the investment instruments widely used by Indonesian society due to its characteristics as a store of value and a hedging asset. However, the price of Antam gold is dynamic because it is influenced by historical price patterns, global gold prices, exchange rates, and domestic macroeconomic conditions. This study aims to forecast Antam gold prices using the XGBoost algorithm by integrating historical price features, technical indicators, and external and macroeconomic variables, including XAUUSD, USDIDR, BI Rate, and inflation. The data used in this study covers the period from 2015 to 2025. The research stages include data collection, data cleaning and alignment, feature engineering, the design of five feature scenarios, hyperparameter tuning using Bayesian Optimization, XGBoost model training, and evaluation using Walk Forward Validation. Model performance was evaluated using RMSE, MAE, and MAPE metrics. The results show that the optimal model performance was not achieved by including all variables, but rather in a specific scenario integrating historical Antam gold price features, technical indicators, XAUUSD, USDIDR, and BI Rate. The best model was obtained using a 60:40 data split and a 7-day test window, resulting in a mean RMSE of 17,170.91, a mean MAE of 15,055.75, and a mean MAPE of 0.959755%. The MAPE value below 1% indicates that the model has a relatively low prediction error during the testing period. Therefore, XGBoost combined with Bayesian Optimization and Walk Forward Validation can be used as a sufficiently accurate approach for forecasting Antam gold prices, particularly for short-term prediction

Item Type:

Thesis (Undergraduate)

Contributors: