Evaluation of Clothing Comfort Using Neural Networks

为了定量评价服装的舒适性，我们对棉、涤纶和改性涤纶进行了穿着试验，并用BP神经网络方法对其服装舒适性进行了评价。从泛化能力的高度证实了神经网络对服装舒适性的精确预测。比较了棉、涤纶和改性涤纶在该系统中泛化能力和权重系数的差异，并利用该神经网络研究了服装面料对服装舒适性的影响。研究还发现，即使是很难准确建模的对象，如磨损试验，神经网络评估也是非常有用的。

To assess clothing comfort quantitatively, we carried out wear trials with cotton, polyester and modified polyester and evaluated their clothing comfort using three-layered neural networks by the back propagation method. It was confirmed that clothing comfort was precisely predictable using neural networks from the height of generalizing ability. The differences among cotton, polyester and modified polyester were observed as the differences in generalizing abilities and weight coefficients in this system and the effects of clothing material on clothing comfort were made dear using this neural networks. It was also found that evaluation by neural networks was very useful for even an object which was difficult to model exactly, such as wear trial.