Peramalan Volatilitas Saham Sektor Pertambangan Emas di Indonesia Berbasis Integrasi Model ARIMA-GARCH-Long Short-Term Memory (LSTM)

Siahaan, Amadea Inchrisa Nauli (2026) Peramalan Volatilitas Saham Sektor Pertambangan Emas di Indonesia Berbasis Integrasi Model ARIMA-GARCH-Long Short-Term Memory (LSTM). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221146-Undergaduate_thesis.pdf - Accepted Version Restricted to Repository staff only Download (18MB) | Request a copy

Abstract

Pasar keuangan modern memiliki karakteristik yang sangat dinamis, di mana volatilitas menjadi indikator krusial dalam menilai risiko investasi. Fluktuasi harga saham, khususnya pada sektor pertambangan emas di Indonesia, sering kali menunjukkan pola heteroskedastisitas dan volatility clustering yang sulit ditangkap secara presisi oleh model linear klasik. Penelitian ini bertujuan untuk memprediksi volatilitas saham sektor pertambangan emas (MDKA.JK, ANTM.JK, BRMS.JK, dan PSAB.JK) menggunakan model integrasi hybrid ARIMA-GARCH-LSTM dengan periode data pengamatan hingga 31 Oktober 2025. Evaluasi performa model dilakukan pada data out-of-sample untuk horizon 1-step, 7-step, dan 20-step ahead berbasis metrik Root Mean Squared Error (RMSE). Hasil pengujian empiris menunjukkan bahwa kinerja model bersifat spesifik kasus (case-specific) dan efektivitas penambahan jaringan non-linier LSTM bergantung pada karakteristik emiten serta panjang horizon peramalan. Model hybrid terbukti secara konsisten mengungguli ARIMA-GARCH konvensional pada horizon jangka pendek (1-step ahead) di seluruh emiten, dengan perbaikan akurasi tertinggi dicapai oleh BRMS.JK (penurunan RMSE sebesar 0,63%) dan PSAB.JK (0,54%). Pada horizon jangka menengah (7-step ahead), model hybrid memberikan peningkatan akurasi paling signifikan pada MDKA.JK dengan pemangkasan RMSE mencapai 14,29%. Sebaliknya, pada horizon jangka panjang (20-step ahead), manfaat model hybrid cenderung berkurang, bahkan mencatatkan kinerja sedikit lebih rendah pada BRMS.JK (penurunan RMSE -0,38%) akibat potensi noise fitting. Proyeksi masa depan (future forecasting) mengindikasikan sensitivitas model hybrid yang tinggi terhadap dinamika informasi terbaru sebagai risk amplifier maupun risk attenuator. Penelitian ini menyimpulkan bahwa model hybrid paling efektif diterapkan pada horizon pendek-menengah dan tidak secara otomatis menjadi pengganti universal bagi model ekonometrika klasik.
===================================================================================================================================
Modern financial markets are highly dynamic, where volatility serves as a crucial indicator for investment risk assessment. Stock price fluctuations, particularly within the Indonesian gold mining sector, often exhibit heteroskedasticity and volatility clustering patterns that are difficult to capture precisely using classical linear models. This study aims to forecast stock volatility in the gold mining sector (MDKA.JK, ANTM.JK, BRMS.JK, and PSAB.JK) using an integrated hybrid ARIMA-GARCH-LSTM model with data observed up to October 31, 2025. Forecasting performance was evaluated on out-of-sample data across 1-step, 7-step, and 20-step ahead horizons using the Root Mean Squared Error (RMSE) metric. The empirical results indicate that model performance is case-specific, and the effectiveness of adding the non-linear LSTM network depends on stock-specific volatility characteristics and the forecasting horizon length. The hybrid model consistently outperformed the conventional ARIMA-GARCH model across all stocks at the short-term horizon (1-step ahead), achieving the highest accuracy improvements in BRMS.JK (0.63% RMSE reduction) and PSAB.JK (0.54% RMSE reduction). At the medium-term horizon (7-step ahead), the hybrid model demonstrated its most substantial performance gain on MDKA.JK, reducing the RMSE by 14.29%. Conversely, for long-term forecasting (20-step ahead), the advantages of the hybrid model diminished, showing a slight performance drop in BRMS.JK (-0.38% RMSE improvement) due to potential noise fitting. True future forecasting projections highlight the high sensitivity of the hybrid model to recent market signals, acting as both a risk amplifier and attenuator. This study concludes that the hybrid model is most beneficial for short-to-medium horizons and should not be treated as a universal replacement for classical econometric models.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	GARCH, Hybrid Model, LSTM, Peramalan, Saham Emas, Volatilitas, Forecasting, GARCH, Gold Mining Stocks, Hybrid Model, LSTM, Volatility
Subjects:	Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry) Q Science > QA Mathematics > QA280 Box-Jenkins forecasting Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Amadea Inchrisa
Date Deposited:	05 Aug 2026 01:55
Last Modified:	05 Aug 2026 01:55
URI:	http://repository.its.ac.id/id/eprint/143840

Actions (login required)

View Item