Analisis Peran Geopolitik Terhadap Brent Crude Price Menggunakan Metode Shapley Additive Explanation Berbasis Ensemble Learning

Alfianur, Muhammad (2026) Analisis Peran Geopolitik Terhadap Brent Crude Price Menggunakan Metode Shapley Additive Explanation Berbasis Ensemble Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006221035-Undergraduate Thesis.pdf - Accepted Version Restricted to Repository staff only Download (12MB) | Request a copy

Abstract

Minyak mentah Brent merupakan komoditas energi vital yang terpengaruh oleh volatilitas yang tinggi dan sensitif terhadap dinamika global. Metode konvensional seperti ARIMA dan GARCH sering kali tidak mampu menangkap pola nonlinear kompleks pada data harga minyak, khususnya saat terjadi guncangan eksternal. Penelitian ini menganalisis kinerja prediksi harga minyak mentah Brent menggunakan tiga algoritma ensemble learning berbasis decision tree, yaitu XGBoost, LightGBM, dan Extra Trees, dengan mengintegrasikan variabel eksogen berupa indeks pasar saham S&P500, Economic Policy Uncertainty (EPU), dan Geopolitical Risk Index (GPR). Data yang digunakan merupakan data harian periode Januari 2005 hingga Desember 2025, dengan pembagian data training dari Januari 2005 hingga Juni 2025 dan data testing dari Juli hingga Desember 2025. Penelitian ini menyusun dua skenario pemodelan yang mana skenario pertama menggunakan 20 fitur lag historis Brent yang kemudian dilakukan seleksi untuk mengambil tiga fitur terbaik menggunakan feature importance berbasis SHAP, dan skenario kedua menambahkan variabel eksogen ke dalam model. Optimasi hyperparameter dilakukan menggunakan grid search, sedangkan evaluasi model menggunakan metrik MAE, RMSE, dan MAPE. Hasil penelitian menunjukkan model terbaik adalah LightGBM pada skenario pertama dengan fitur lag 5, 11, dan 13, menghasilkan MAE sebesar 1,662, RMSE sebesar 2,291, dan MAPE sebesar 2,589%. Analisis SHAP pada skenario kedua mengungkapkan bahwa fitur S&P500 menjadi kontributor dominan pada ketiga model dengan proporsi kontribusi mencapai 40,5% pada XGBoost, 44% pada LightGBM, dan 69,3% pada Extra Trees, diikuti oleh EPU dan GPR. Hasil penelitian ini memberikan wawasan bagi pemangku kepentingan industri energi dalam mempertimbangkan faktor sentimen pasar dan risiko geopolitik global untuk mitigasi risiko harga minyak mentah.

====================================================================================================================================================================================

Brent crude oil is a vital energy commodity that is affected by high volatility and sensitive to global dynamics. Conventional methods such as ARIMA and GARCH are often unable to capture the complex nonlinear patterns in oil price data, particularly during external shocks. This study analyzes the predictive performance of Brent crude oil prices using three decision tree-based ensemble learning algorithms XGBoost, LightGBM, and Extra Trees by incorporating exogenous variables including the S&P500 stock market index, Economic Policy Uncertainty (EPU), and Geopolitical Risk Index (GPR). The dataset consists of daily observations from January 2005 to December 2025, with training data spanning January 2005 to June 2025 and testing data covering July to December 2025. This study develops two modeling scenarios, in which the first scenario employs 20 historical lag features of Brent crude that are subsequently selected to retain the top three features using SHAP-based feature importance, while the second scenario incorporates exogenous variables into the model. Hyperparameter optimization was conducted using grid search, and model performance was evaluated using MAE, RMSE, and MAPE metrics. Results indicate that LightGBM under the first scenario, utilizing lag features 5, 11, and 13, achieved the best performance with an MAE of 1.662, RMSE of 2.291, and MAPE of 2.589%. SHAP analysis in the second scenario revealed that the S&P500 feature was the dominant contributor across all three models, accounting for 40.5% in XGBoost, 44% in LightGBM, and 69.3% in Extra Trees, followed by EPU and GPR. The findings of this study provide insights for energy industry stakeholders in considering market sentiment and global geopolitical risk factors for crude oil price risk mitigation.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Brent Crude, Ensemble Learning, Geopolitik, SHAP, Volatilitas Brent Crude, Ensemble Learning, Geopolitical, SHAP, Volatility
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Muhammad Alfianur
Date Deposited:	16 Jul 2026 02:59
Last Modified:	16 Jul 2026 02:59
URI:	http://repository.its.ac.id/id/eprint/135121

Actions (login required)

View Item