Nonlinear, Non-stationary and Seasonal Time Series Forecasting Using Different Methods Coupled with Data Preprocessing
Procedia Computer Science 2017
Artūrs Stepčenko, Jurijs Čižovs, Ludmila Aleksejeva, Juri Tolujew

Time series forecasting is important in several applied domains because it facilitates decision-making in this domains. Commonly, statistical methods such as regression analysis and Markov chains, or artificial intelligent methods such as artificial neural networks (ANN) are used in forecasting tasks. In this paper different time series forecasting methods were compared using the normalized difference vegetation index (NDVI) time series forecasting. NDVI is a nonlinear, non-stationary and seasonal time series used for short-term vegetation forecasting and management of various problems, such as prediction of spread of forest fire and forest disease. In order to reduce input data set dimensionality and improve predictability, stepwise regression analysis and principal component analysis (PCA) were used as data pre-processing techniques. For comparing the obtained performance for the different methods, several performance criteria commonly used in forecasting statistical evaluation were calculated.


Keywords
Artificial neural networks, Markov chains, Principal component analysis, Ridge regression, Stepwise regression
DOI
10.1016/j.procs.2017.01.175
Hyperlink
http://www.sciencedirect.com/science/article/pii/S187705091730176X

Stepčenko, A., Čižovs, J., Aleksejeva, L., Tolujew, J. Nonlinear, Non-stationary and Seasonal Time Series Forecasting Using Different Methods Coupled with Data Preprocessing. Procedia Computer Science, 2017, Vol. 104, pp.578-585. ISSN 1877-0509. Available from: doi:10.1016/j.procs.2017.01.175

Publication language
English (en)
The Scientific Library of the Riga Technical University.
E-mail: uzzinas@rtu.lv; Phone: +371 28399196