Support vector machines for predicting worsted yarn properties

Abstract

Support vector machines (SVMs) models have been presented for predicting worsted yarn properties using SVM regression algorithms. Model selection which amounts to search in hyper-parameter space is performed to study the suitable parameter conditions. The predictive powers of the SVM models have been estimated and the results are compared with ANN models. It is observed that under the small population circumstances, SVM models are still capable of maintaining the stability of predictive accuracy, and more suitable for noisy and dynamic spinning process.