Predicting the qualitative characteristics of yarns, such as tensile properties, unevenness, and hairiness from the raw material properties, has been the main purpose of many studies in recent decades. In this study, a total of forty different rovings obtained from carded and combed short-staple cotton (for increasing variation limits of fiber properties) were selected from various spinning mills. In the first section, the feasibility of two-way prediction of cotton yarn properties and fiber properties using multivariate methods has been presented. Multivariate multiple regression was used to predict the dependent variables. There are reasonable agreements between the estimated properties with experimental values. Then, the optimal models for the dependent variables included yarn imperfection were obtained. In the second section, we investigated some statistical approaches for modeling and prediction of important properties of ring spun cotton yarn. The derived models indicated better prediction performance than the previous studies. We also selected yarn imperfection as a dependent variable. . In the final section, because it might be that the observed/obtained data and relations among variables are not crisp, the fuzzy least squares regression was used for the estimation of yarn properties. The results indicated that all of yarn properties (strength, elongation, unevenness, and hairiness of yarn) are influenced by fiber tenacity, short fiber index, fineness of fibers, roving unevenness, and yarn count. Most of parameters in the equations are obtained as fuzzy numbers. We also explain the obtained optimal equation, for each dependent variable as follows:
