An approach to represent time series forecasting via fuzzy numbers

Atakan Sahin, Tufan Kumbasar, Engin Yesil, M. Furkan Dodurka, Onur Karasakal

Research output: Chapter in Book/Report/Conference proceedingConference contribution book

2 Citations (Scopus)
63 Downloads (Pure)

Abstract

This paper introduces a new approach for estimating the uncertainty in the forecast through the construction of Triangular Fuzzy Numbers (TFNs). The interval of the proposed TFN presentation is generated from a Fuzzy logic based Lower and Upper Bound Estimator (FLUBE). Here, instead of the representing the forecast with a crisp value with a Prediction Interval (PI), the level of uncertainty associated with the point forecasts will be quantified by defining TFNs (linguistic terms) within the uncertainty interval provided by the FLUBE. This will give the opportunity to handle the forecast as linguistic terms which will increase the interpretability. Moreover, the proposed approach will provide valuable information about the accuracy of the forecast by providing a relative membership degree. The demonstrated results indicate that the proposed FLUBE based TFN representation is an efficient and useful approach to represent the uncertainty and the quality of the forecast.

Original languageEnglish
Title of host publicationProceedings: Second International Conference on Artificial Intelligence, Modelling, and Simulation, AIMS 2014
Place of PublicationPiscataway, NJ.
PublisherIEEE
Pages51-56
Number of pages6
ISBN (Print)9781479975990
DOIs
Publication statusPublished - 5 May 2015
Event2nd IEEE International Conference on Artificial Intelligence, Modelling, and Simulation, AIMS 2014 - Madrid, Spain
Duration: 18 Nov 201420 Nov 2014

Conference

Conference2nd IEEE International Conference on Artificial Intelligence, Modelling, and Simulation, AIMS 2014
CountrySpain
CityMadrid
Period18/11/1420/11/14

Keywords

  • forecasting
  • fuzzy estimator
  • fuzzy numbers
  • fuzzy time series

Fingerprint Dive into the research topics of 'An approach to represent time series forecasting via fuzzy numbers'. Together they form a unique fingerprint.

Cite this