Rancang Bangun Dashboard Peramalan Harga Minyak Mentah Menggunakan Pendekatan Long Short-Term Memory Berbasis Attention Mechanism

Afriansyah, Moreno Reyhan (2026) Rancang Bangun Dashboard Peramalan Harga Minyak Mentah Menggunakan Pendekatan Long Short-Term Memory Berbasis Attention Mechanism. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2043221050-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Minyak mentah merupakan komoditas energi fundamental yang krusial menggerakkan roda perekonomian global. Perannya sebagai bahan baku utama bahan bakar dan berbagai produk industry membuat harga minyak sangat fluktuatif, dipengaruhi oleh kondisi makroekonomi seperti US Dollar (USDX) dan risiko geopolitik dunia (Geopolitical Risk Index atau GPR). Sebagai negara importir neto, volatilitas harga minyak menjadi ancaman serius bagi stabilitas ekonomi nasional Indonesia. Oleh karena itu, kebutuhan peramalan yang akurat menjadi urgensi bagi pemerintah dalam memproyeksikan nilai impor di masa depan. Penelitian ini menerapkan metode Deep Learning, khususnya Long Short-Term Memory (LSTM) berbasis Attention Mechanism untuk menangkap pola kompleksitas pada data deret waktu. Kombinasi metode ini dirancang agar mampu menangani dependensi jangka panjang serta memberikan atensi lebih pada peristiwa geopolitik dunia dan fluktuasi USDX guna meningkatkan akurasi peramalan. Hasil penelitian menunjukkan bahwa model terbaik yang diperoleh melalui Ablation Study pada case 4 dengan menggunakan fitur West Texas Intermediate (WTI) Lag 1 dan GPR. Model tersebut menghasilkan nilai Mean Absolute Percentage Error (MAPE) sebesar 2,606% pada data testing yang mengindikasikan tingkat akurasi peramalan yang sangat tinggi. Hasil peramalan harga minyak mentah WTI selama 30 hari ke depan yaitu 5 Agustus 2025 hingga 3 September 2025 menunjukkan adanya pola tren positif.
==============================================================================================================================================
Crude oil is a fundamental energy commodity that plays a crucial role in driving the global economy. Its role as a primary raw material for fuels and various industrial products causes oil prices to be highly volatile, influenced by macroeconomic conditions such as the US Dollar Index (USDX) and global geopolitical risks (Geopolitical Risk Index or GPR). As a net importing country, oil price volatility poses a serious threat to Indonesia's national economic stability. Therefore, the need for accurate forecasting becomes a pressing concern for the government in projecting future import values. This study applies Deep Learning methods, specifically Long Short-Term Memory (LSTM) based on the Attention Mechanism, to capture the complex patterns in time series data. This combination of methods is designed to handle long-term dependencies and provide greater attention to global geopolitical events and USDX fluctuations in order to enhance forecasting accuracy. The study results indicate that the best model, obtained through an Ablation Study in Case 4 using WTI Lag 1 and GPR features, achieved a Mean Absolute Percentage Error (MAPE) of 2.606% on the testing data, reflecting a very high level of forecasting accuracy. The forecast for WTI crude oil prices over the next 30 days, from August 5, 2025, to September 3, 2025, demonstrates a positive trend pattern.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Attention Mechanism, Geopolitical Risk Index, LSTM, Minyak Mentah, Peramalan Attention Mechanism, Crude Oil, Forecasting, Geopolitical Risk Index, LSTM
Subjects:	T Technology > T Technology (General) > T174 Technological forecasting
Divisions:	Faculty of Vocational > 49501-Business Statistics
Depositing User:	Moreno Reyhan Afriansyah
Date Deposited:	02 Feb 2026 03:41
Last Modified:	02 Feb 2026 03:41
URI:	http://repository.its.ac.id/id/eprint/131554

Actions (login required)

View Item