A COMPARISON OF ARTIFICIAL NEURAL NETWORKS AND MULTIPLE LINEAR REGRESSION MODELS AS IN PREDICTORS OF FABRIC WEFT DEFECTS


Kargi V. S.

TEKSTIL VE KONFEKSIYON, vol.24, no.3, pp.309-316, 2014 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 24 Issue: 3
  • Publication Date: 2014
  • Journal Name: TEKSTIL VE KONFEKSIYON
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, TR DİZİN (ULAKBİM)
  • Page Numbers: pp.309-316
  • Keywords: Artificial neural network, Multilayer perceptron model, Multiple linear regression model, Fabric weft defect, Prediction
  • Bursa Uludag University Affiliated: Yes

Abstract

Predicting uncertainty is quite important for the reliability of decisions to be made by business managers. Contemporary problems are complex, and their solutions require scientific decision-making. The aim of this study is to predict weft defects in fabric production for a textile business using a multilayer perceptron model and multiple linear regression models. Matlab R2010b software was used for multilayer perceptron model solutions, and SPSS 13 packet software was used for multiple linear regression model solutions. The results of the two models were compared, and the multilayer perceptron model was identified as the best predictive model. This study shows that in operational research both artificial neural networks and the multiple linear regression model can be successfully used to predict fabric weft errors.