Implementasi Geopolitical Risk Index Terhadap Prediksi Volatilitas Bitcoin Menggunakan Metode Hybrid GARCH-X-SVR

Iskartama, Intan Salsabilla (2025) Implementasi Geopolitical Risk Index Terhadap Prediksi Volatilitas Bitcoin Menggunakan Metode Hybrid GARCH-X-SVR. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006211090-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Pasar keuangan idealnya mencerminkan seluruh informasi yang tersedia secara efisien, tetapi pasar cryptocurrency, seperti Bitcoin, sering kali menunjukkan ketidakefisienan yang ditandai dengan tingginya volatilitas harga. Bitcoin, sebagai mata uang kripto pertama dan paling dominan, memiliki karakteristik desentralisasi dan transparansi yang menarik bagi investor, tetapi juga penuh risiko. Volatilitas Bitcoin tidak hanya dipengaruhi oleh faktor internal, seperti harga masa lalu, tetapi juga oleh faktor eksternal, salah satunya adalah risiko geopolitik yang diukur menggunakan Geopolitical Risk Index (GPR). Penelitian ini bertujuan untuk menganalisis pengaruh GPR Index terhadap volatilitas return Bitcoin serta meningkatkan akurasi prediksi dengan menggabungkan model GARCH dengan metode SVR dalam pendekatan hybrid GARCH-X-SVR. Model GARCH-X digunakan karena mampu menangkap volatilitas yang terpengaruh oleh variabel eksogen, sementara SVR untuk mengoreksi hasil prediksi GARCH dan meningkatkan akurasi prediksi melalui pendekatan machine learning. Hasil penelitian dengan menggunakan metode GARCH-X menunjukkan bahwa GPR Index berpengaruh signifikan terhadap volatilitas return Bitcoin sebesar 0,009132. Integrasi SVR dengan kernel RBF ke dalam model GARCH dan GARCH-X terbukti meningkatkan akurasi prediksi. Model GARCH-SVR dengan kernel RBF menurunkan nilai evaluasi MSE sebesar 53,06% dan QLIKE sebesar 50,99% dibanding GARCH tradisional. Sementara itu, model GARCH-X-SVR menunjukkan penurunan MSE sebesar 53,76% dan QLIKE sebesar 50,23% dibandingkan GARCH-X, menandakan bahwa integrasi SVR dan GPR Index mampu meningkatkan kemampuan model dalam menangkap dinamika volatilitas, terutama saat volatilitas tinggi. Penambahan GPR Index dalam model GARCH-X-SVR meningkatkan performa prediksi volatilitas return Bitcoin, ditunjukkan dengan penurunan nilai MSE sebesar 1,17% dibandingkan model GARCH-SVR, tetapi, nilai evaluasi QLIKE model GARCH-X-SVR 1,39% lebih besar dibanding model GARCH-SVR. Meskipun penambahan GPR Index memberikan informasi tambahan, evaluasi trade-off antara kedua metrik evaluasi menunjukkan bahwa model GARCH-SVR sendiri sudah cukup akurat dalam memprediksi volatilitas Bitcoin, tanpa perlu penambahan variabel eksogen.
=====================================================================================================================================
Financial markets are ideally expected to reflect all available information efficiently, yet the cryptocurrency market, particularly Bitcoin, often demonstrates inefficiencies marked by high price volatility. Bitcoin, as the first and most dominant cryptocurrency, possesses decentralized and transparent characteristics that attract investors but also carry significant risks. Bitcoin's volatility is influenced not only by internal factors such as past prices but also by external factors, one of which is geopolitical risk measured using the Geopolitical Risk Index (GPR). This study aims to analyze the effect of the GPR Index on Bitcoin return volatility and to enhance prediction accuracy by combining the GARCH model with the Support Vector Regression (SVR) method in a hybrid GARCH-X-SVR approach. The GARCH-X model is utilized for its ability to capture volatility influenced by exogenous variables, while SVR serves to correct GARCH predictions and improve accuracy through a machine learning approach. The results from the GARCH-X method indicate that the GPR Index has a significant effect on Bitcoin return volatility, with a coefficient of 0,009132. The integration of SVR with the RBF kernel into the GARCH and GARCH-X models significantly improves prediction accuracy. The GARCH-SVR model with RBF kernel reduces the MSE by 53,06% and QLIKE by 50,99% compared to the traditional GARCH model. Meanwhile, the GARCH-X-SVR model shows a decrease in MSE by 53,76% and QLIKE by 50,23% compared to GARCH-X, indicating that the integration of SVR and the GPR Index enhances the model’s ability to capture volatility dynamics, especially during periods of high volatility. The addition of the GPR Index in the GARCH-X-SVR model improved the prediction performance of Bitcoin return volatility, shown by a 1,17% reduction in MSE compared to the GARCH-SVR model. However, the QLIKE evaluation value of the GARCH-X-SVR model was 1,39% higher than the GARCH-SVR model. Although the inclusion of the GPR Index provides additional information, the trade-off evaluation between the two metrics indicates that the GARCH-SVR model alone is already sufficiently accurate in predicting Bitcoin volatility without the need for exogenous variables.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Bitcoin, GARCH, Geopolitical Risk Index, SVR, Volatilitas, Bitcoin, GARCH, Geopolitical Risk Index, SVR, Volatility
Subjects:	Q Science > Q Science (General) > Q180.55.M38 Mathematical models Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry) Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression Q Science > QA Mathematics > QA353.K47 Kernel functions (analysis)
Divisions:	Faculty of Mathematics, Computation, and Data Science > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Intan Salsabilla Iskartama
Date Deposited:	25 Jul 2025 07:59
Last Modified:	25 Jul 2025 07:59
URI:	http://repository.its.ac.id/id/eprint/121502

Actions (login required)

View Item