Abstract: We investigate 36 kinds of knitted fabrics and their thermal-wet comfort objective evaluation indices and their subjective in-wear evaluation indices. Based on the least squares support vector machines (LSSVM), we create regression models to predict the subjective evaluation using objective evaluation indices as input. We systematically evaluate the learned regression model using statistical learning methods. Moreover, we compare the regression model based on LSSVM with the regression model based on the back propagation neural network (BP). According to the experimental results, the LSSVM model yields more accurate predictions on subjective thermal-wet evaluations than the BP model.