Andayuri, Naufal Raihan (2024) Penerapan Metode Hybrid Autoregressive Integrated Moving Average - Support Vector Regression (ARIMA-SVR) dalam Peramalan Harga Bitcoin. Other thesis, Institut Teknologi Sepuluh Nopember.
Text
5003201110-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy |
Abstract
Pada era saat ini, kemajuan teknologi berkembang pesat, termasuk di sektor ekonomi dan keuangan. Salah satu inovasi penting dalam sektor keuangan adalah cryptocurrency seperti Bitcoin. Teknologi blockchain yang terkait dengan Bitcoin telah membawa transparansi yang tinggi dalam pelacakan pembayaran digital. Bitcoin, meskipun memiliki risiko tinggi, juga memiliki potensi keuntungan yang besar jika dikelola dengan baik. Karena jumlah Bitcoin yang terbatas, meskipun nilai tukarnya fluktuatif, nilai Bitcoin cenderung meningkat dari waktu ke waktu. Oleh karena itu, analisis terhadap fluktuasi harga Bitcoin penting untuk mengurangi risiko investasi. Tujuan dari penelitian ini yaitu untuk memodelkan dan melakukan peramalan pada harga Bitcoin. Permodelan ARIMA digunakan untuk menangkap pola linear dalam data. Permodelan ARIMA bisa jadi menghasilkan residual yang belum white noise, sehingga residual tersebut bisa dimodelkan dengan SVR. Kedua metode tersebut digabungkan sehingga menghasilkan model Hybrid ARIMA-SVR. Keakuratan dan kesalahan peramalan dari ketiga metode ini dibandingkan menggunakan RMSE dan MAPE. Hasil analisis menunjukkan bahwa model ARIMA(1,1,1) untuk data harga Bitcoin menghasilkan RMSE sebesar 850,92 dan MAPE sebesar 1,559% pada data testing. Pemodelan harga Bitcoin dengan SVR menggunakan kernel RBF memberikan hasil optimum dengan RMSE sebesar 841,14 dan MAPE sebesar 1,516%. Model hybrid ARIMA-SVR dengan ARIMA(1,1,0) yang tidak memenuhi asumsi white noise menghasilkan RMSE sebesar 832,90 dan MAPE sebesar 1,510%. Model hybrid ARIMA-SVR terbukti menjadi yang terbaik untuk meramalkan harga Bitcoin selama 7 hari mendatang karena memiliki nilai RMSE dan MAPE paling kecil. Secara keseluruhan, pemodelan harga Bitcoin menggunakan ARIMA, SVR, dan hybrid ARIMA-SVR menunjukkan hasil yang akurat berdasarkan nilai error MAPE dan RMSE. Namun, masalah time lag yang terjadi harus diperhatikan karena meskipun tingkat akurasi model baik, prediksi seringkali terlambat.
========================================================================================================================
In the current era, technological advances are growing rapidly, including in the economic and financial sectors. One of the notable innovations in the financial sector is cryptocurrencies such as Bitcoin. Blockchain technology associated with Bitcoin has brought a high degree of transparency in digital payment tracking. Bitcoin, despite its high risk, also has the potential for huge profits if managed properly. Due to the limited number of Bitcoins, even though the exchange rate fluctuates, the value of Bitcoin tends to increase over time. Therefore, analysis of Bitcoin price fluctuations is important to reduce investment risk. The purpose of this study is to model and forecast the price of Bitcoin. ARIMA modeling is used to capture linear patterns in data. ARIMA modeling may produce residues that are not yet white noise, so that the residuals can be modeled with SVR. The two methods are combined to produce the Hybrid ARIMA-SVR model. The accuracy and forecasting errors of these three methods are compared using RMSE and MAPE. The results of the analysis show that the ARIMA(1,1,1) model for Bitcoin price data produces an RMSE of 850.92 and a MAPE of 1.559% in the testing data. Bitcoin price modeling with SVR using the RBF kernel gave optimal results with an RMSE of 841.14 and a MAPE of 1.516%. The ARIMA-SVR hybrid model with ARIMA(1,1,0) that does not meet the white noise assumption produces an RMSE of 832.90 and a MAPE of 1.510%. The ARIMA-SVR hybrid model proved to be the best for forecasting the price of Bitcoin over the next 7 days as it has the least RMSE and MAPE values. Overall, Bitcoin price modeling using ARIMA, SVR, and hybrid ARIMA-SVR shows accurate results based on MAPE and RMSE error values. However, the time lag problem that occurs must be considered because although the accuracy level of the model is good, the predictions are often late.
Item Type: | Thesis (Other) |
---|---|
Uncontrolled Keywords: | ARIMA, Bitcoin Price, Forecasting, Hybrid ARIMA-SVR, SVR, ARIMA, Harga Bitcoin, Hybrid ARIMA-SVR, Peramalan, SVR. |
Subjects: | H Social Sciences > HA Statistics > HA30.3 Time-series analysis H Social Sciences > HB Economic Theory > Economic forecasting--Mathematical models. |
Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis |
Depositing User: | Naufal Raihan Andayuri |
Date Deposited: | 08 Aug 2024 12:04 |
Last Modified: | 08 Aug 2024 12:04 |
URI: | http://repository.its.ac.id/id/eprint/114898 |
Actions (login required)
View Item |