基于最小方差支持向量机的织物热湿舒适性预测

Prediction of fabric thermal-moisture comfort based on least squares support vector machines

在纺织服装工程研究中应用人工智能与机器学习的方法,可以更加准确地预测纺织材料的穿着热湿舒适性。为此,利用最小方差支持向量机(LSSVM),分析了36种针织织物热湿舒适性客观指标与人体穿着对织物的热湿舒适性主观评定之间的对应关系,并建立了客观指标与主观评定之间的回归模型。该模型能够快速预测成衣之后人体穿着主观评定的舒适度,并可节约新面料和织物材料研发过程中的评估成本。通过对多个回归模型的比较与分析,证明LSSVM回归模型比BP神经网络模型能够更加准确地预测织物的主观热湿舒适性。

The application of artificial intelligence and machine learning methods to textile fashion engineering facilitates the prediction of thermal-moisture comfort of fabrics.Thirty-six kinds of knitted fabrics are investigated and the relationship between the thermal-moisture comfort objective evaluation indices and the subjective wear evaluation indices of the fabrics are analyzed by least squares support vector machines（LSSVM）.And regression models are created to predict the subjective evaluation using objective evaluation indices as input parameters.These models can quickly predict the subjective wear comfort when the fabrics are made into clothes and significantly reduce the cost of evaluation for the development of new fabric materials.Analytical comparison of different regression models demonstrated that the LSSVM model yields more accurate prediction of subjective thermal-moisture comfort than the BP model.