Perbandingan Model Multivariate Time Series, Backpropagation Neural Network, dan Model Hybrid dalam Peramalan Harga Saham Sektor Emas Indonesia Berdasarkan Harga Emas Dunia

Aulia, Taqy (2025) Perbandingan Model Multivariate Time Series, Backpropagation Neural Network, dan Model Hybrid dalam Peramalan Harga Saham Sektor Emas Indonesia Berdasarkan Harga Emas Dunia. Diploma thesis, Institut Teknologi Sepuluh Nopember.

Text 5006221076-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Peningkatan harga emas dunia yang mencapai level tinggi dalam beberapa periode terakhir serta perannya sebagai aset lindung nilai (safe haven) di tengah ketidakpastian ekonomi global meningkatkan perhatian investor terhadap saham perusahaan pertambangan emas di Indonesia. Pergerakan saham sektor emas yang volatil dan dipengaruhi fluktuasi harga emas dunia memerlukan metode yang mampu menangkap hubungan linier sekaligus pola nonlinier pada data time series. Penelitian ini bertujuan menganalisis hubungan antara harga emas dunia (XAUUSD) dan harga penutupan saham sektor emas di Indonesia, yaitu PT Aneka Tambang Tbk (ANTM), PT Bumi Resources Minerals Tbk (BRMS), dan PT J Resources Asia Pasifik Tbk (PSAB), serta membandingkan kinerja metode peramalan Vector Autoregression (VAR), Backpropagation Neural Network (BPNN), dan model hybrid VAR–BPNN. Data yang digunakan berupa data time series harian periode 1 Januari 2020 hingga 31 Desember 2025 yang diperoleh dari Investing.com. Analisis meliputi pengujian stasioneritas dan kointegrasi, pemodelan VAR, Impulse Response Function (IRF), Forecast Error Variance Decomposition (FEVD), pemodelan BPNN, dan forecasting hybrid VAR–BPNN. Hasil penelitian menunjukkan bahwa tidak terdapat hubungan kointegrasi jangka panjang antarvariabel sehingga digunakan model VAR(1). Hasil IRF dan FEVD menunjukkan bahwa harga emas dunia memiliki pengaruh jangka pendek yang cukup penting terhadap saham sektor emas, terutama ANTM. Pada peramalan, model BPNN memberikan performa terbaik dibandingkan VAR(1) dan hybrid VAR–BPNN pada seluruh variabel penelitian dengan tingkat kesalahan prediksi terendah, sedangkan model hybrid belum mampu meningkatkan akurasi prediksi secara signifikan karena residual VAR diduga belum mengandung pola nonlinier yang cukup kuat.
===================================================================================================================================
The increase in global gold prices reaching high levels in recent periods, along with its role as a safe-haven asset amid global economic uncertainty, has increased investor interest in gold mining company stocks in Indonesia. The volatile movement of gold sector stocks and their sensitivity to global gold price fluctuations require methods capable of capturing both linear and nonlinear patterns in time series data. This study aims to analyze the relationship between global gold prices (XAUUSD) and the closing prices of Indonesian gold sector stocks, namely PT Aneka Tambang Tbk (ANTM), PT Bumi Resources Minerals Tbk (BRMS), and PT J Resources Asia Pasifik Tbk (PSAB), as well as to compare the forecasting performance of Vector Autoregression (VAR), Backpropagation Neural Network (BPNN), and the hybrid VAR–BPNN model. This study uses daily time series data from January 1, 2020 to December 31, 2025 obtained from Investing.com. The analysis includes stationarity and cointegration testing, VAR modeling, Impulse Response Function (IRF), Forecast Error Variance Decomposition (FEVD), BPNN modeling, and hybrid VAR–BPNN forecasting. The results indicate no long-term cointegration relationship among variables; therefore, a VAR(1) model is employed. IRF and FEVD results show that global gold prices have an important short-term influence on gold mining stocks, particularly ANTM. In forecasting performance, BPNN outperformed VAR(1) and hybrid VAR–BPNN across all variables by producing the lowest prediction errors, while the hybrid model did not significantly improve forecasting accuracy because VAR residuals were presumed not to contain sufficiently strong nonlinear patterns.

Item Type:	Thesis (Diploma)
Uncontrolled Keywords:	Backpropagation Neural Network, Harga Emas Dunia, Hybrid VECM–BPNN, Saham Sektor Emas, Vector Autoregression
Subjects:	Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Taqy Aulia
Date Deposited:	24 Jul 2026 02:57
Last Modified:	24 Jul 2026 02:57
URI:	http://repository.its.ac.id/id/eprint/136572

Actions (login required)

View Item