Loading...
Please wait, while we are loading the content...
Similar Documents
Predicción de series de tiempo no lineales usando un modelo híbrido basado en la descomposición wavelet
| Content Provider | Semantic Scholar |
|---|---|
| Author | Londoño, Michael Vásquez |
| Copyright Year | 2019 |
| Abstract | The forecast of nonlinear time series has received special attention due to the difficulty to capture the underlying patterns in the data. In recent years, a number of models have been proposed to improve the performance of the traditional models. In this research, we used a methodology to forecast nonstationary and nonlinear time series contaminated with high noise levels. In particular, we combine the maximal overlap discrete wavelet transform (MODWT) with the model ARFIMA-HYGARCH and neural networks (NN). In addition, the LSW model is implemented to forecast non-stationary time series. In the present study, both methodologies are applied to forecast USD/COP exchange rate. The results suggest that the hybrid methodology, based on wavelets and neural networks, provided more accurate forecast of an appreciation or a depreciation of the exchange rates. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://repositorio.unal.edu.co/bitstream/handle/unal/64765/TesisMsc_Michael.pdf?isAllowed=y&sequence=1 |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
