Peramalan Harga Nikel Global Menggunakan Metode Hybrid Generalized Autoregressive Conditional Heteroscedasticity-Long Short-Term Memory (GARCH-LSTM)

Hariyadi, Nadhifa Putri (2025) Peramalan Harga Nikel Global Menggunakan Metode Hybrid Generalized Autoregressive Conditional Heteroscedasticity-Long Short-Term Memory (GARCH-LSTM). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006211108-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Peramalan harga nikel global menjadi sangat penting mengingat meningkatnya permintaan terhadap komoditas ini, terutama untuk produksi baterai kendaraan listrik (EV) serta berbagai aplikasi industri. Pesatnya pertumbuhan pasar EV, seiring dengan upaya global dalam transisi menuju energi terbarukan, mendorong kebutuhan akan model prediksi harga yang akurat dan andal. Harga nikel bersifat sangat fluktuatif, dipengaruhi oleh ketegangan geopolitik, gangguan rantai pasok, serta dinamika permintaan industri yang terus berubah, sehingga metode prediksi tradisional sering kali tidak memadai. Penelitian ini mengusulkan model prediksi hibrida yang mengintegrasikan Generalized Autoregressive Conditional Heteroscedasticity (GARCH) dan jaringan saraf Long Short-Term Memory (LSTM) untuk mengatasi tantangan tersebut. Model GARCH mampu menangkap volatilitas jangka pendek dan tren linear pada data harga, sementara LSTM efektif dalam mengidentifikasi pola non-linear yang kompleks serta ketergantungan jangka panjang yang kerap terabaikan oleh model konvensional. Analisis dilakukan terhadap data harga nikel global dari Januari 2021 hingga Januari 2025, periode yang mencerminkan ketidakstabilan pasar akibat pemulihan ekonomi pasca-pandemi, konflik geopolitik, serta perubahan kebijakan perdagangan. Gabungan GARCH dan LSTM menunjukkan kinerja prediktif unggul dengan Mean Absolute Percentage Error (MAPE) sebesar 2,98%, menandakan tingkat akurasi yang tinggi. Model ini terbukti adaptif terhadap guncangan pasar mendadak dan tren yang terus berkembang, menjadikannya alat yang bermanfaat bagi investor, pembuat kebijakan, serta pelaku industri. Hasil utama penelitian mengindikasikan bahwa pendekatan hibrida menghasilkan prediksi tujuh hari ke depan yang lebih stabil dibandingkan model tunggal, dengan harga yang diproyeksikan menunjukkan tren kenaikan bertahap dalam kisaran 15.737,61 hingga 16.131,77 USD dan deviasi yang minimal. Studi ini tidak hanya memajukan metodologi prediksi di sektor komoditas, tetapi juga memberikan kontribusi praktis dalam pengambilan keputusan strategis. Selain itu, kerangka kerja yang ditawarkan memiliki potensi untuk diterapkan pada komoditas lain yang juga memiliki karakter volatil, seperti litium dan kobalt, sehingga dapat meningkatkan prediktabilitas pasar serta mendukung perencanaan ekonomi berkelanjutan. =====================================================================================================================================
Forecasting global nickel prices has become increasingly important due to the rising demand for this commodity, particularly for the production of electric vehicle (EV) batteries and various industrial applications. The rapid growth of the EV market, alongside global efforts to transition toward renewable energy, has driven the need for accurate and reliable price prediction models. Nickel prices are highly volatile, influenced by geopolitical tensions, supply chain disruptions, and the ever-changing dynamics of industrial demand, making traditional forecasting methods often inadequate. This study proposes a hybrid predictive model that integrates Generalized Autoregressive Conditional Heteroscedasticity (GARCH) and Long Short-Term Memory (LSTM) neural networks to address these challenges. The GARCH model effectively captures short-term volatility and linear trends in price data, while the LSTM component is capable of identifying complex non-linear patterns and long-term dependencies that are often overlooked by conventional models. The analysis is conducted using global nickel price data from January 2021 to January 2025, a period marked by market instability due to post-pandemic economic recovery, geopolitical conflicts, and shifts in trade policy. The combination of GARCH and LSTM demonstrates superior predictive performance, achieving a Mean Absolute Percentage Error (MAPE) of 2.98%, indicating a high level of accuracy. This model proves to be adaptive to sudden market shocks and evolving trends, making it a valuable tool for investors, policymakers, and industry practitioners. The main findings of this study indicate that the hybrid approach delivers more stable seven-day-ahead forecasts compared to single models, with projected prices showing a gradual upward trend ranging between USD 15,737.61 and USD 16,131.77, accompanied by minimal deviations. This study not only advances forecasting methodology in the commodity sector but also offers practical insights for strategic decision-making. Moreover, the proposed framework holds potential for application to other volatile commodities such as lithium and cobalt, thereby enhancing market predictability and supporting sustainable economic planning.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	GARCH, Hybrid model, LSTM, Model hybrid, Nickel price forecasting, Peramalan harga nikel, Volatilitas, Volatility.
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Nadhifa Putri Hariyadi
Date Deposited:	31 Jul 2025 03:50
Last Modified:	31 Jul 2025 03:50
URI:	http://repository.its.ac.id/id/eprint/119803

Actions (login required)

View Item