Title: Framework to identify a set of univariate time series forecasting techniques to aid in business decision making

Authors: Iram Naim; Tripti Mahara

Addresses: Department of Polymer and Process Engineering, IIT, Roorkee, U.K., India ' Department of Polymer and Process Engineering, IIT, Roorkee, U.K., India

Abstract: Forecasting is generally involved in business activities to anticipate or predict the future. With availability of numerous techniques and models, forecasters regularly face a genuine issue to identify suitable technique for different time series available in an organisation. Most of the time, it is not possible to find one technique that can be used for all-time series as the selection is dependent upon the characteristics of a time series. Hence, the research proposes a selection tree to aid in decision making based upon availability of type of dataset and time series characteristics. The framework is validated using four real case studies. This study also presents advancement to existing forecasting method selection tree by exploring a new dimension of complex seasonal pattern for long time series.

Keywords: univariate time series; time series pattern; model selection; trend analysis; seasonal data; complex seasonality; long time series.

DOI: 10.1504/IJIE.2020.110764

International Journal of Intelligent Enterprise, 2020 Vol.7 No.4, pp.423 - 443

Received: 17 Nov 2018
Accepted: 24 Feb 2019

Published online: 01 Sep 2020 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article