Institutional Repository of Vilnius University
Forecasting nonstationary and nearly nonstationary time series using machine learning methods /
Abstract
dc:descriptionThe purpose of the study is to examine whether the differences between nonstationary and nearly nonstationary time series have a significant impact on forecasting using machine learning models. One of the main objectives is to verify the ability of the machine learning models to make quite accurate one-step ahead predictions for both nonstationary and nearly nonstationary time series. Another objective is to compare the forecasting performance of machine learning and traditional statistical models. For the analysis three machine learning models, including the Multilayer Perceptron (MLP) network, Recurrent Neural Network (RNN) and Support Vector Regression (SVR), and a single traditional statistical method Autoregressive Integrated Moving Average (ARIMA) are used. This work consists of the simulation study, including the first order autoregressive time series case analysis, and the application to the real world data with an example of financial market time series. The results show that in most cases machine learning models predict both nonstationary and nearly nonstationary time series quite accurately. However, machine learning models are not able to make significantly better predictions for the time series, which follow a random walk, in comparison to the traditional statistical methods.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Vaičiūnaitė, Reda,
- Contributors dc:contributor
-
- Markevičiūtė, Jurgita
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- eng
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD81590231&prefLang=en_US