基于支持向量机的精纺毛织物透气性预测

Prediction of the worsted fabrics′ air permeability based on support vector machine

为了能够快速、高效地预测精纺毛织物材料透气性,首先对精纺毛织物透气性参数之间的关系进行分析,然后,基于对现有的精纺毛织物透气性预测方法进行比较分析,并借助支持向量机训练速度快、参数选择少的优点,构建了一种基于支持向量机的精纺毛织物透气性预测模型。其次,选取了34组精纺毛织物样本,其中27组作为训练样本,7组作为测试样本,通过提出的模型对精纺毛织物的透气性进行实验验证。结果表明:在参数C=1 325.525 8和σ=0.102 8的条件下,对精纺毛织物透气性预测结果的平均误差小于4%,得到较好预测结果。与现有BP神经网络预测模型相比,其预测精度提高了3%,进一步说明构建的模型有利于快速、高效地预测精纺毛织物材料透气性。

To quickly and efficiently predict the air permeability of the worsted fabrics,the relationship between the worsted fabrics′parameters and the air permeability was deeply analyzed.After compared and analyzed the existing forecasting methods,and took use of the advantages of fast training and less parameter selection with SVM,a regression model of worsted fabrics′air permeability based on support vector machine was presented in this paper,and the relationship between the 34 kinds of worsted fabrics′parameters and the air permeability was analyzed.In the 34 groups of samples,27 groups were randomly selected as training sets,and 7 groups were used as test sets.Under the condition of C=1 325.525 8 andσ=0.102 8,the experiment achieved good results,and the average prediction accuracy was as low as 4%.Compared with the BP regression model,the SVM model reduced the average experimental error by 3%.This paper demonstrated that the SVM model yields more accurate prediction of worsted fabrics′air permeability than the BP model.