2000 journal article
Fabric softness classification using linear and nonlinear models
TEXTILE RESEARCH JOURNAL, 70(3), 201–204.
In this study, the authors use linear and nonlinear models and yarn parameters such as CV%, hairiness, and surface softness to classify the softness of knitted fabrics (T-shirts) for comparison to human subjective evaluations. All classification rates are verified with a leave-one-out cross-validation technique. The results show 20% misclassification when using a linear model to sort samples into two classes (low and high). When sorting into three classes, the misclassification is 30%. When sorting T-shirt softness into three classes using a tree modeling technique and the surface response average (SRA) and maximum peak-to-valley height (Ry), it is possible to match the human data at a 65% rate. When using surface response parameters and measured yam properties to sort T-shirt softness into three classes, with tree modeling it is possible to classify 91% of the samples accurately based on the human data.