Peramalan Harga Minyak Mentah Dunia dengan Menggunakan Long Short-Term Memory (LSTM)

Theresia, Laura Wilhelmina (2024) Peramalan Harga Minyak Mentah Dunia dengan Menggunakan Long Short-Term Memory (LSTM). Other thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 5026201137-Undergraduate_Thesis.pdf] Text
5026201137-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 October 2026.

Download (3MB) | Request a copy

Abstract

Minyak mentah memiliki peran krusial dalam perekonomian global, mempengaruhi aktivitas konsumsi dan produksi di setiap negara. Fluktuasi harga minyak mentah secara signifikan berdampak pada stabilitas ekonomi, termasuk di Indonesia. Ketidakstabilan harga minyak dapat menimbulkan kesulitan dalam mengelola kebutuhan dalam negeri dan menyusun Anggaran Pendapatan dan Belanja Negara (APBN). Maka dari itu, peramalan harga minyak mentah menjadi krusial untuk membantu pemerintah mengambil keputusan yang tepat dalam merespons perubahan pasar. Tugas akhir ini menggunakan metode Long Short-Term Memory (LSTM), sebuah jenis rekurensi dalam jaringan saraf yang dikenal efektif dalam memodelkan hubungan temporal dalam data time series. LSTM digunakan untuk meramalkan harga minyak mentah WTI dengan pendekatan univariat dan multivariat. Hasil penelitian menunjukkan bahwa model LSTM terbaik adalah model dengan proporsi data 60%:40% dan data univariat, yang menghasilkan nilai Mean Absolute Percentage Error (MAPE) sebesar 1,76%. Parameter terbaik yang digunakan adalah20 epochs, batch size 16, dan 64 unit LSTM dengan sequence length 1. Model LSTM menunjukkan akurasi yang lebih baik dibandingkan model Radial Basis Function Neural Network (RBFNN) dan Extreme Learning Machine (ELM), meskipun membutuhkan waktu eksekusi yang lebih lama. Hasil peramalan juga menunjukkan bahwa proporsi data dan normalisasi mempengaruhi akurasi dan kecepatan eksekusi model. Model LSTM memberikan hasil peramalan yang konsisten dan akurat dengan MAPE rendah pada kedua jenis data, univariat dan multivariat. Namun, peramalan multivariat tidak selalu menghasilkan akurasi yang lebih baik dibandingkan univariat, menunjukkan bahwa variabilitas data dan pengaturan parameter model memiliki pengaruh signifikan terhadap kualitas peramalan.
=================================================================================================================================
Crude oil plays a crucial role in the global economy, influencing consumption and production activities in every country. Fluctuations in crude oil prices significantly impact economic stability, including in Indonesia. Price instability can create difficulties in managing domestic
needs and formulating the State Budget (APBN). Therefore, forecasting crude oil prices is crucial to help the government make informed decisions in response to market changes. This final project utilizes the Long Short-Term Memory (LSTM) method, a type of recurrent neural network known for effectively modeling temporal relationships in time series data. LSTM is used to forecast WTI crude oil prices using univariate and multivariate approaches. The study results indicate that the best LSTM model is the one with a 60%:40% data proportion and univariate data, producing a Mean Absolute Percentage Error (MAPE) of 1.76%. The best parameters used are 20 epochs, batch size of 16, and 64 LSTM units with a sequence length of1. The LSTM model shows better accuracy compared to the Radial Basis Function Neural Network (RBFNN) and Extreme Learning Machine (ELM) models, despite requiring longer execution times. The forecasting results also show that data proportion and normalization affect the model's accuracy and execution speed. The LSTM model provides consistent and accurate forecasting results with low MAPE for both univariate and multivariate data. However, multivariate forecasting does not always yield better accuracy than univariate forecasting, indicating that data variability and model parameter settings significantly influence forecasting quality.

Item Type: Thesis (Other)
Uncontrolled Keywords: Jaringan Saraf, Long Short-Term Memory (LSTM), Minyak Mentah, Peramalan Time Series, Crude Oil, neural network, time series forecasting.
Subjects: T Technology > T Technology (General) > T174 Technological forecasting
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User: Laura Wilhelmina Theresia
Date Deposited: 12 Jul 2024 08:58
Last Modified: 12 Jul 2024 08:58
URI: http://repository.its.ac.id/id/eprint/108279

Actions (login required)

View Item View Item