Size fitting problem is a main obstacle to large scale online garment sales.It is the difficult to customers to find the fit garments when they couldn't try on. IntroductionWith the development of shopping online, the status of garment network sales is rising gradually. According to the statistics, 50% returns are because of unfitted problems since customers can't try garments on by online sales [1]. Many researches devote themselves to find a method can keep customer's measurements match with garment size. In garment size selection algorithm, least-square method has been applied on garment size classification for a long time [2]-[4], it can develop the rationality by using weighting method. Yu [5] put forward a garment size selection method by comparing the matching relationship between figures and garment size. Through the comparison between chest、 、 waist hip width of figures and the ones on garments, define the matching relationship between figures and garment size. He [6] made research on Yu's method, and proposed modified advice. Xu put forward using AHP method to select garment size [7],the method makes size selection into quantification, which is suitable for garment sales online. On the basis of AHP, Ding and Xu also developed a garment size selection method combined immune algorithms and AHP [8], and support that the prediction accuracy of it is higher than using AHP alone. The method from Chen, made use of BP network to process size classification [9], can apply on garment size selection the same. Some garment size matching algorithms are completed in the form of virtual fitting. In this case, garment pattern has been regarded as a pure geometry problem transferred from two dimensions into three dimensions. Structure grid was made on the patterns, and then finite element method is used to calculate the surface curvature of garment models in order to process salient points matching automatically [10]- [11]. Propose a co-evolutionary immune algorithm for the multicriteria decision making (MCDM) model, and use the model to solve the large scale garment matching problem [12]. The application of neural network on size selection use the measurement points as input and the corresponding garment sizes as output to train the network [13]-[14] .In this paper, we developed the method of garment size matching on the basis of wavelet neural network. It could input new somatic data and actual suitable garment size by trying on in the follow-up step, by means of training reduce the discrepancy between neural network output and actual results.

No abstract