A novel procedure for multimodel development using the grey silhouette coefficient for small-data-set forecasting

Che Jung Chang, Wen Li Dai, Chien Chih Chen

Research output: Contribution to journalArticlepeer-review

11 Citations (Scopus)

Abstract

Small-data-set forecasting problems are a critical issue in various fields, with the early stage of a manufacturing system being a good example. Manufacturers require sufficient knowledge to minimize overall production costs, but this is difficult to achieve due to limited number of samples available at such times. This research was thus conducted to develop a modelling procedure to assist managers or decision makers in acquiring stable prediction results from small data sets. The proposed method is a two-stage procedure. First, we assessed some single models to determine whether the tendency of a real sequence can be reflected using grey incidence analysis, and we then evaluated their forecasting stability based on the relative ratio of error range. Second, a grey silhouette coefficient was developed to create an applicable hybrid forecasting model for small samples. Two real cases were analysed to confirm the effectiveness and practical value of the proposed method. The empirical results showed that the multimodel procedure can minimize forecasting errors and improve forecasting results with limited data. Consequently, the proposed procedure is considered a feasible tool for small-data-set forecasting problems.

Original languageEnglish
Pages (from-to)1887-1894
Number of pages8
JournalJournal of the Operational Research Society
Volume66
Issue number11
DOIs
Publication statusPublished - 2015 Nov 1

All Science Journal Classification (ASJC) codes

  • Management Information Systems
  • Strategy and Management
  • Management Science and Operations Research
  • Marketing

Fingerprint Dive into the research topics of 'A novel procedure for multimodel development using the grey silhouette coefficient for small-data-set forecasting'. Together they form a unique fingerprint.

Cite this