Construction of Small Scale Hierarchy Neural Network for Clothing Comfort Evaluation of Skirt

Abstract

In this thesis, a small-scale hierarchy neural network (HNN), which can forecast the clothing comfort of trial skirts using some simple information at an apparel department, is described.
The input information to HNN is “Individual information” and “Product information”, and is input by the consumer or a salesperson. The individual information is a consumer's figure data that consists of five items. The content of the product information is the material of the trial skirts and the design. Final HNN is a small-scale network that is constructed with an input layer includes 11 units, one hidden layer includes 12 units and an output layer indudes three units. The HNN forecasted clothing comfort of nine kinds of trial skirts. As a result, the agreement rate with data used training has been 90% or more, and with data submitted to network has been from 62 to 70%.
It is thought that this small scale HNN is an effective forecast method, because though the clothing comfort is composed of complex factors, constructed HNN can forecast the clothing comfort by using information to which consumer or salesperson can treat easily.