Peramalan Harga Nikel Indonesia dengan Model Transformer

Adriani, Ovi Wahyu (2026) Peramalan Harga Nikel Indonesia dengan Model Transformer. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2043221062-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Perkembangan teknologi yang mendorong peralihan menuju energi ramah lingkungan membuat produsen kendaraan bermotor saling berlomba mengembangkan kendaraan listrik. Salah satu material utama yang banyak dimanfaatkan dalam produksi baterai kendaraan listrik adalah nikel. Peramalan harga nikel menjadi hal penting bagi perusahaan dalam menetapkan kebijakan, seperti penentuan harga jual dan perencanaan pembelian bahan baku. Peramalan harga nikel di Indonesia menghadapi tantangan tersendiri karena pergerakan harganya yang cenderung fluktuatif dan kompleks. Kondisi tersebut menyulitkan pihak-pihak terkait, seperti perusahaan, investor, dan konsumen, dalam mengambil keputusan bisnis, operasional, maupun konsumsi secara tepat. Penelitian ini menerapkan model Transformer yang dikenal memiliki kinerja unggul dalam meramalkan data nonlinier dibandingkan metode peramalan konvensional seperti ARIMA dan neural network. Tujuan utama penelitian ini adalah untuk mengevaluasi kinerja model Transformer dalam meramalkan harga nikel di Indonesia serta memperoleh hasil prediksi yang akurat. Data yang digunakan merupakan data historis harga acuan nikel periode Oktober 2017 hingga Agustus 2025. Hasil penelitian menunjukkan bahwa model Transformer terbaik dihasilkan melalui skenario 2 pembagian data 90:10 dengan konfigurasi arsitektur yang terdiri dari 2 encoder layers, 4 multi-head attention, dimensi embedding sebesar 16, dan feed-forward dimension sebesar 32. Hal ini dibuktikan dengan perolehan nilai metrik evaluasi paling minimal, yaitu Mean Absolute Percentage Error (MAPE) sebesar 3,68% dan Mean Absolute Error (MAE) sebesar 551,72 USD/dmt. Hasil peramalan harga nikel Indonesia untuk 24 bulan ke depan (September 2025 – Agustus 2027) lebih naik dari data aktual, dengan nilai hasil peramalan pada periode September 2025 sebesar 15352,17 USD/dmt dan Agustus 2027 sebesar 16032,87 USD/dmt.
===================================================================================================================================
Technological developments driving the transition toward environmentally friendly energy have led motor vehicle manufacturers to compete in developing electric vehicles (EVs). Nickel is one of the primary materials extensively utilized in the production of electric vehicle batteries. Consequently, forecasting nickel prices is crucial for companies in establishing policies, such as determining selling prices and planning raw material procurement. Nickel price forecasting in Indonesia faces unique challenges due to price movements that tend to be volatile and complex. Such conditions make it difficult for relevant stakeholders, including companies, investors, and consumers, to make accurate business, operational, and consumption decisions. This study implements the Transformer model, which is recognized for its superior performance in forecasting non-linear data compared to conventional methods such as ARIMA and standard neural networks. The primary objective of this research is to evaluate the performance of the Transformer model in forecasting Indonesian nickel prices and to obtain accurate prediction results. The dataset used consists of historical nickel reference prices from October 2017 to August 2025. The results indicate that the optimal Transformer model was achieved through Scenario 2 with a 90:10 data split, utilizing an architectural configuration of 2 encoder layers, 4 multi-head attention heads, an embedding dimension of 16, and a feed-forward dimension of 32. This is evidenced by the lowest evaluation metric values, specifically a Mean Absolute Percentage Error (MAPE) of 3.68% and a Mean Absolute Error (MAE) of 551.72 USD/dmt. The forecasting results for Indonesian nickel prices over the next 24 months (September 2025 – August 2027) show an upward trend compared to actual data, with forecasted values of 15,352.17 USD/dmt in September 2025 and 16,032.87 USD/dmt in August 2027.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Harga Nikel Indonesia, MAE, MAPE, Model Transformer, Peramalan, Indonesian Nickel Price, MAE, MAPE, Transformer Model, Forecasting
Subjects:	Q Science Q Science > Q Science (General) Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Vocational > 49501-Business Statistics
Depositing User:	Ovi Wahyu Adriani
Date Deposited:	10 Feb 2026 03:54
Last Modified:	10 Feb 2026 03:54
URI:	http://repository.its.ac.id/id/eprint/132282

Actions (login required)

View Item