Research On Fashion Color Quantification And Prediction

Fashion color prediction is to predict the future color trend of fashion in24months byspecial methods and to provide trend palettes. Obtained palettes could offer clothingenterprises valuable information on fashion color during the future selling seasons, which isbeneficial for drafting manufacturing and marketing plots. However, the lag and confidence ofpalettes during propagating restrict its commercial value and stimulate the development ofresearch on fashion color prediction. Though several theories and methods have been appliedinto this field and achievements are obtained, fashion color prediction is still in exploringstage. Deficiencies and controversies existed and the validity of prediction has already been apressing problem of the day in the textile and clothing industries.Color is with typically sensitive simulation and fashion color with ambiguous expression.How to translate the sensitive color into rational data and build serious mathematicalprediction models are the key points of fashion color prediction. In this research, the methodof color quantification was discussed, based on which the prediction researches on fashioncolor were investigated. Fashion color palettes were taken as the objects. The palettes werereleased by International Commission for Color Fashion and Textiles (INTERCOLOR) forwomen’s spring/summer, from2007to2013. Quantification and classification methods oncolors were put forward. The process was based on the analyses of PANTONE color system,which was applied by the palettes. Then features of palettes on hue, lightness and chromawere obtained respectively analyzing the statistical data. Consequently, hue prediction modelswere established using grey system theory and back-propagation neural networks respectively.The predicted effects of the models were investigated by setting different time series. Finally,lightness and chroma prediction models were established on the basis of hue. The systematicprediction on fashion color was developed.The innovations of this thesis are illustrated as follows:1) Quantification andclassification of color based on PANTONE color system were proposed in terms of thenon-standard and disunity of color quantification in fashion color prediction researches. It canprovide a valuable reference for the standardization of color and improve the communicationbetween colors on subjective perception and the objective computation application.2) Thepredicted effects of the prediction models were investigated by setting different time series,which can provide soundly basis for fashion color prediction researches.3) Prediction methodon lightness and chroma of color was put forward based on the dominant hues, which enrichesthe theory on fashion color research. The main studies of this paper were arranged as follows.Preface of the thesis was illustrated in Chapter1. Research background, methodologyand literature was introduced. Research levels and insufficiencies in this field are summarized.Consequently, the purpose, significance and methods of this research were proposed.Quantification and analyses of fashion color palettes were discussed in Chapter2.Inercolor palettes for women’s spring/summer, from2007to2013, were proposed as objectsand PANTONE color system was as the quantification tool, against the problems of disunityof data resources and complexity of quantification methods. Quantification and classification criteria of hue, lightness and chroma were defined. Consequently, features of palettes wereobtained by analyzing statistical data of hue, lightness and chroma.The predictions of hues using grey system theory were investigated in Chapter3.Prediction models on hue were established by grey model, in accordance with the significanceand the nonlinear features of hue. Validities of the prediction models were discussed bysetting different time series of historical fashion color data, as4,5and6years respectively.Mean absolute error (MAE) was considered as evaluation indicator in this thesis. Resultsdemonstrated that GM (1,1) cannot predict the hues continuously due to the high requirementon historical data and limit the generality of the prediction.The predictions on hue by using back-propagation neural networks (BPNN) wereprobed in Chapter4, in accordance with the deficiency of grey system theory. The validity ofthe prediction models were explored systematically by designing different structures of BPNNwith input neurons as3,4and5respectively.Consequently, prediction on lightness and chroma were investigated in Chapter5and6.Prediction models on mean values of lightness and chroma, based on dominant hues, wereproposed against the absence of internal relation among the three attributes of colors.Finally, the main achievements of the research were summed up in Chapter7of thethesis, meanwhile problems to be solved and explored in the future researches were pointedout.Quantification and classification methods were provided based on fashion color palettesin this thesis. Fashion color prediction was discussed by using grey system theory and BPNN,which was based on analyses of statistical data of fashion color palettes. It conducted theprediction with a higher accuracy and the prediction results could provide textile industriesreference on future fashion color trend.