A FCM, Grey Model, and BP Neural Network Hybrid Fashion Color Forecasting Method

Abstract

In view of the low prediction accuracy of the existing fashion color prediction methods, this paper propose a fashion color forecasting method used the spring and summer women’s fashion color data released by the International Fashion Color Committee from 2007 to 2013. In preprocess stage, the Pantone color system is used as the color quantization basis, the fuzzy c-means is used to cluster the sample data at first, and a FCM algorithm is used to statistic the color categories in different time series. In forecasting stage, both the grey model and BP neural network are used respectively to construct the fashion color hue prediction model from the statistical results generated from FCM. In evaluation stage, the mean square error is used to compare the prediction effect. The results show that the grey model based on FCM has the smallest error and has the best prediction effect. The proposed model can be used to predict the future fashion color, which can help the apparel industry stakeholders to grasp the trend of the future fashion color and make design and production plan more effectively. The FCM and grey model hybrid prediction method shown in this model also can be used in other small sample data prediction scenario.