Markov-Switching GARCH-Machine Learning Untuk Meramalkan Volatilitas Return Sukuk (Studi Kasus: Franklin Global Sukuk Fund Luxemburg)

Yani, Dinda Ulima Rizky (2023) Markov-Switching GARCH-Machine Learning Untuk Meramalkan Volatilitas Return Sukuk (Studi Kasus: Franklin Global Sukuk Fund Luxemburg). Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6003201008-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2025. Download (6MB) | Request a copy

Abstract

Variasi faktor risiko dan volatilitas dari return merupakan kunci untuk menentukan masalah alokasi portofolio investasi. Prediksi volatilitas yang akurat menjadi hal yang penting untuk menentukan nilai aset secara efektif. Namun dalam jangka panjang terdapat banyak yang hal dapat menyebabkan terjadinya perubahan perilaku dalam data keuangan yang menyebabkan adanya perubahan struktural pada volatilitas dan model tipe GARCH standar gagal menangkap pergerakan volatilitas tersebut. Selain itu, model ekonometrik sederhana dilandasi oleh asumsi-asumsi tertentu yang harus dipenuhi. Untuk itu diperlukan sebuah analisis yang dapat menangkap pergeseran regime volatilitas tanpa dibatasi oleh asumi-asumsi tertentu. Pada penelitian ini dilakukan peramalan volatilitas menggunakan metode MSGARCH-Machine Learning pada return Franklin Global Sukuk Fund Luxemburg dimana metode Machine Learning (ML) yang digunakan meliputi Feedforward Neural Network (FFNN), Support Vector Regression (SVR) dan Long Short Term Memory (LSTM). MSGARCH digunakan untuk memodelkan volatilitas yang mengalami pergeseran regime dimana setiap regime diasumsikan berdistribusi normal. Sedangkan metode-metode ML digunakan untuk memodelkan volatilitas return berdasarkan volatilitas masing-masing regime tanpa dilandasi asumsi distribusi tertentu. Hasil peramalan antara model MSGARCH-ML dan model MSGARCH standar dibandingkan untuk mengetahui performa masing-masing metode. Akurasi dari model diukur menggunakan MSE dan sMAPE. Model MSGARCH-FFNN dan MSGARCH-SVR menghasilkan nilai error yang lebih kecil dari model MSGARCH standar, namun dengan struktur input yang berbeda. Namun tidak ada satupun model yang mengkombinasikan MSGARCH dan LSTM yang lebih baik dari model MSGARCH standar. Kombinasi metode MSGARCH dan machine learning (MSGARCH-ML) mendapatkan hasil peramalan yang lebih baik dari MSGARCH standar tergantung pada metode machine learning apa yang digunakan dan bagaimana struktur input yang digunakan
=====================================================================================================================================
Variations in risk and volatility of returns are the keys to determining investment portfolio allocation problems. Accurate volatility prediction is important to determine asset value effectively. However, there are many things that can cause behavioral changes in the long term financial data that lead to structural changes in volatility and the standard GARCH type model fails to capture these volatility movements. In addition, simple econometric models are built based on certain assumptions. For this reason, an analysis that can capture the shift in the volatility regime without being limited by certain assumptions is needed. In this study, volatility forecasting was carried out using the MSGARCH – Machine Learning, such as Feedforward Neural Network (FFNN), Support Vector Regression (SVR) and Long Short Term Memory (LSTM), on the return of Franklin Global Sukuk Fund Luxemburg. MSGARCH is used to model the volatility of shifting regimes where each regime is assumed to be normally distributed. Machine Learning (ML) methods can be used to model return volatility based on the volatility of each regime without any particular assumption. Furthermore, the forecasting results between the MSGARCH-ML model and the standard MSGARCH model will be compared to determine the performance of each method. The accuracy of the model was measured using the MSE and sMAPE. The MSGARCH-FFNN and MSGARCH-SVR models produce smaller error values than the MSGARCH standard model, but with different input structures. However, no model that combines MSGARCH and LSTM is better than the standard MSGARCH model. The combination of the MSGARCH and machine learning methods (MSGARCH-ML) gets better forecasting results than the MSGARCH standard depending on what machine learning method is used and how the input structure is used

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	volatility, return, sukuk, MSGARCH, machine learning, volatilitas, return, sukuk, MSGARCH, machine learning
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) > Statistics > 49101-(S2) Master Thesis
Depositing User:	Dinda Ulima Rizky Yani
Date Deposited:	13 Feb 2023 03:58
Last Modified:	13 Feb 2023 03:58
URI:	http://repository.its.ac.id/id/eprint/97135

Actions (login required)

View Item