The Research On The Forecast Of Women's SS Fashion Color Based On BP Neural Network

With the prevailing of color economy, more and more attention has been paid on the added value brought by fashion color in textile and garment industry, color forecasting has become an indispensable link. Most of the domestic color theories are from abroad because of the late start of China's apparel industry, foreign authorities releases play an important role as reference and domestic enterprises are dependent on the color information released by such institutions as International Fashion Color Association, Cotton Incorporated, Pantone Company and so on. Today, based on the combination of domestic market conditions, more and more domestic institutions can do a comprehensive prediction job in reference to foreign release. However, in the era of the highly developed computer technology, the existing prediction methods at home and abroad still reveal flaws, such as practicality, reachability and the problems of handling the nonlinear variation of fashion color.This subject searches and summarizes the domestic and international relevant literatures on basic theories of fashion color characteristic rules, color technology, forecast and application in order to have an overview of fashion color and its forecasts, and then explore the changing rules from the perspective of subjective and objective, analyzes the advantages and disadvantages of the forecasting thoughts and impacting forecasting factors for the preparation of quantitative forecast. As the BP neural network can approach any relationship function through artificial intelligence methods, complete self-programming design according to the need and then output data that are closer to the final result, this subject discusses the scientific, comprehensive, continuous characters of this method. Firstly, Women's SS fashion color cards from 2001 to 2011 released by authoritative organ are collected and quantified with a scientific method of CIELab color measurement, then, the color characteristic data are classified as the object of experimental study, at last, fashion color hue , tone and the tone of single color are forecasted and estimated by the establishment of sample learning database and training model of the BP neural network.This subject clearly illustrate the vivid fashion color characteristics via color cards and quantitative data charts, which shows that there is triple features of women's SS fashion color, constant, gradation and mutation, this provides foundation for quantitative fashion color forecast. Also fashion color forecast not only has a support of favorable conditions like color quantization, fashion color rules, but also received negative influences like abrupt changes, objective restrictive, subjective uncertainty factors. In order to maximize the key parameters of the decision project, fashion color hue and tone number, the tone distribution of red color corresponding to the min and max value of L*, a* and b*, are forecasted and estimated in 2011 and re-forecasted in 2012 by the use of the BP neural network. Forecast results in 2011 are basically coincident with actual fashion trends, which shows a scientific, comprehensive, continuous fashion color forecast is done and color forecast theories are enriched.