Analisis Risiko dan Peramalan Harga Bitcoin dan Emas Menggunakan Metode Long Short Term Memory (LSTM)

Kirana, Raditya Septa (2025) Analisis Risiko dan Peramalan Harga Bitcoin dan Emas Menggunakan Metode Long Short Term Memory (LSTM). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006211038-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Investasi menjadi salah satu kegiatan yang digemari masyarakat karena dapat meningkatkan keuntungan dari kegiatan tersebut, Perkembangan teknologi dan digitalisasi telah mendorong berkembangnya alat investasi seperti Bitcoin, serta aset tradisional seperti emas, sebagai instrumen investasi. Namun, volatilitas harga kedua aset tersebut menghadirkan tantangan dalam memprediksi pergerakan harga dan mengukur risiko yang terkait. Penelitian ini bertujuan untuk menganalisis risiko dan memprediksi harga Bitcoin dan emas menggunakan pendekatan model Long Short-Term Memory (LSTM), salah satu metode dalam jaringan saraf tiruan yang terbukti efektif dalam menangani data deret waktu. Proses penelitian dimulai dengan pengumpulan data harga harian Bitcoin dan emas dalam periode tahun 2019-2024. Data dianalisis melalui tahapan preprocessing, pengujian stasioneritas, normalisasi, dan pemodelan LSTM. Kinerja model dievaluasi menggunakan indikator Mean Absolute Percentage Error (MAPE). Selain itu, risiko investasi diukur dengan pendekatan simulasi Monte Carlo untuk menghitung nilai VaR masing-masing asset. Hasil penelitian menunjukkan model LSTM terbaik untuk saham emas dengan kombinasi hyperparameter yaitu dengan 50 epoch, 128 unit, dan dropout sebesar 0.2 menghasilkan performa terbaik dengan nilai MAPE sebesar 1.064%. sedangkan untuk model LSTM pada saham bitcoin yaitu menggunakan kombinasi 50 epoch, 128 unit, dan dropout 0.2 menunjukkan performa terbaik dengan nilai MAPE terendah sebesar 1.979%. Prediksi ini menunjukkan kecenderungan harga yang meningkat pada kedua aset. Estimasi VaR mengindikasikan bahwa bitcoin memiliki risiko kerugian lebih tinggi dibandingkan emas.
=================================================================================================================================
Investment has become one of the activities favored by the public because it can increase profits from such activities. The development of technology and digitalization has driven the emergence of investment tools like Bitcoin, as well as traditional assets like gold, as investment instruments. However, the price volatility of both assets presents challenges in predicting price movements and measuring the associated risks. This research aims to analyze the risk and predict the prices of Bitcoin and gold using the Long Short-Term Memory (LSTM) model approach, one of the methods in artificial neural networks that has proven effective in handling time series data. The research process begins with the collection of daily price data for Bitcoin and gold during the period from 2019 to 2024. Data is analyzed through preprocessing, stationarity testing, normalization, and LSTM modeling stages. The model's performance is evaluated using the Mean Absolute Percentage Error (MAPE). In addition, investment risk is measured using the Monte Carlo simulation approach to calculate the VaR value of each asset. The research results show that the best LSTM model for gold stocks, with a combination of hyperparameters of 50 epochs, 128 units, and a dropout of 0.2, produced the best performance with an MAPE value of 1.064%. whereas for the LSTM model on Bitcoin stocks, using a combination of 100 epochs, 128 units, and a dropout of 0.2 shows the best performance with the lowest MAPE value of 1.979%. This prediction shows an upward price trend for both assets. The VaR estimate indicates that Bitcoin has a higher risk of loss compared to gold.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Bitcoin, Emas, Long Short Therm Memory (LSTM), Analisis risiko, Prediksi harga, Gold, Long Short-Term Memory (LSTM), Risk analysis, Price prediction
Subjects:	Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Raditya Septa Kirana
Date Deposited:	29 Jul 2025 02:05
Last Modified:	29 Jul 2025 02:05
URI:	http://repository.its.ac.id/id/eprint/122003

Actions (login required)

View Item