Determining the yarn quality by feature selection and classification in a yarn production facility

41116

Nowadays, computer technology is rapidly advancing, computer capacities are increasing, which makes it easier to reach the database by increasing the number of information recording areas. However, when the produced and recorded data are meaningless on their own, they become meaningful when processed for a certain purpose. Converting raw data to meaningful information can be done by data mining. In this study, rule extraction is realized in a yarn production facility by classification which is one of the data mining methods. Prior to classification, the features that affect the yarn quality are determined, and feature selection is realized by choosing the effective features by Taguchi experimental design method. Rule extraction phase is applied for both cost-insensitive classification that aims to minimize the number of misclassification errors and cost-sensitive classification that aims to minimize the expected misclassification cost. For rule extraction Weka 3.8.1 and MT-VeMa 1.0 package programs are used. The resulting rules guide the firm for producing qualified yarns. This study presents how data mining and experimental design applications at a textile firm have been achieved with actual data and the contributions to the processes of the firm.

713

Feature selection , Taguchi experimental design , Data mining , Classification

Pamukkale University Journal Of Engineering Sciences

719

Pamukkale University Journal Of Engineering Sciences