Saputri, Prilyandari Dina (2017) Aplikasi Model Hybrid Quantile Regression Neural Network pada Peramalan Pecahan Inflow dan Outflow Uang Kartal di Indonesia. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
1313100020-Undergraduate_Theses.pdf - Published Version Download (3MB) | Preview |
Abstract
Regresi kuantil merupakan perluasan dari regresi Ordinary Least Square (OLS) yang dapat menjelaskan keterkaitan antar variabel pada berbagai kuantil. Regresi kuantil juga dapat digunakan dalam peramalan data runtun waktu. Penelitian ini bertujuan untuk memperoleh model peramalan inflow dan outflow tiap pecahan di Indonesia yang akurat dan dapat menangkap pola variasi kalender serta heteroskedastisitas. Untuk meningkatkan akurasi hasil peramalan, regresi kuantil dikombinasikan dengan neural network, yang dikenal sebagai quantile regression neural network (QRNN). Metode QRNN akan dibandingkan dengan metode ARIMAX dan neural network berdasarkan RMSE, MAE, MdAE, MAPE, dan MdAPE. Terdapat dua kajian dalam penelitian ini, yakni studi simulasi dan aplikasi pada 14 pecahan data inflow dan outflow di Indonesia. Studi simulasi menunjukkan bahwa QRNN dapat menangkap pola heteroskedastisitas dan nonlinieritas dibandingkan ARIMAX dan neural network. Sedangkan aplikasi pada data inflow dan outflow menunjukkan bahwa QRNN merupakan metode terbaik dalam meramalkan 10 dari 14 pecahan. Namun peramalan interval metode QRNN menunjukkan adanya crossing antar kuantil yang disebabkan oleh pengestimasian kuantil secara independen.
=================================================================
Quantile regression was developed from Ordinary Least Square regression. Furthermore, quantile regression can explain the relationship between variables on various quantiles. Quantile regression can be applied in forecasting analysis. The aim of this study was to find the best model for forecasting inflow and outflow in Indonesia which can overcome heteroscedasticity and nonlinearity problem. In order to improve the accuracy of forecasting results, quantile regression will be combined with neural network method, known as quantile regression neural network (QRNN). QRNN will be compared with ARIMAX and neural network method based on RMSE, MAE, MdAE, MAPE, and MdAPE criteria. In this study, there are two main topics will be discussed, i.e simulation study and case study about 14 currencies of inflow and outflow data. Simulation study shows that QRNN is the best method to solve heteroscedasticity and nonlinearity problem. While, application in inflow and outflow data shows that QRNN is the best method to forecast 10 of 14 currencies. However, there is a crossing within quantile which can be caused by the estimates of each quantile that are calculated independently.
| Item Type: | Thesis (Undergraduate) |
|---|---|
| Additional Information: | RSSt 519.536 Sap a |
| Uncontrolled Keywords: | Inflow dan Outflow, Heteroskedastisitas, Neural Network, Nonlinieritas, Regresi Kuantil |
| Subjects: | H Social Sciences > HB Economic Theory > Economic forecasting--Mathematical models. H Social Sciences > HG Finance Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression |
| Divisions: | Faculty of Mathematics and Science > Statistics > 49201-(S1) Undergraduate Thesis |
| Depositing User: | Prilyandari Dina Saputri |
| Date Deposited: | 25 Oct 2017 03:15 |
| Last Modified: | 05 Mar 2019 04:04 |
| URI: | http://repository.its.ac.id/id/eprint/48490 |
Actions (login required)
 |
View Item |