Suitable Models for Seasonal and Trend Time Series Forecasting

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26510%2F15%3APU116615" target="_blank" >RIV/00216305:26510/15:PU116615 - isvavai.cz</a>

Result on the web

<a href="http://www.konference.fbm.vutbr.cz/ic/" target="_blank" >http://www.konference.fbm.vutbr.cz/ic/</a>

DOI - Digital Object Identifier

—

Result language

angličtina

Original language name

Suitable Models for Seasonal and Trend Time Series Forecasting

Original language description

Purpose of the article is to compare suitable method for time series forecasting, especially the method of sea-sonal and trend time series decomposition and neural network prediction and use of methods defined by Box and Jenkins (Box-Jenkins models), such as ARIMA or SARIMA. For the analysis data from the area of wholesale trade with connecting material is used. Methodology/methods used in the paper consists of time series analysis, such as seasonal and trend decomposi-tion using time series adjustment from the effect of calendar variations, decomposition of multiplicative time-series model, prediction with neural networks and Box-Jenkins autoregressive integrated moving average mod-els. Last but not least it is worth noting deductive quantitative methods for research and data analysis using graphs. Scientific aim is to compare the effectiveness of traditional statistical models with artificial neural networks models. Autoregressive integrated moving average model is recently very popular linear method for time series prediction. Last research activities in forecasting with artificial neural networks show that the combination of time series decomposition and further prediction with artificial neural network can also be a suitable method for this purpose. Findings of the research show that artificial neural networks models can be a promising alternative to the tradi-tional linear models. Conclusions (limits, implications etc) resulting from the paper are beneficial for further research. The conducted research suggests methods of time series analysis and decomposition, artificial neural networks and Box-Jenkins models are suitable instruments for seasonal and trend time series forecasting. The article presents selected methods as very useful and bringing many opportunities for further research.

Czech name

—

Czech description

—

Type

D - Article in proceedings

CEP classification

—

OECD FORD branch

50602 - Public administration

Project

—

Continuities

S - Specificky vyzkum na vysokych skolach

Publication year

2015

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection

Perspectives of Business and Entrepreneurship Development

ISBN

978-80-214-5227-5

ISSN

—

e-ISSN

—

Number of pages

8

Pages from-to

422-429

Publisher name

Faculty of Business and Management, Brno University of Technology, 2015

Place of publication

Brno

Event location

Brno

Event date

May 28, 2015

Type of event by nationality

WRD - Celosvětová akce

UT code for WoS article

000428947700042