基于灰色前馈神经网络的流行色预测

提出结合灰色理论和前馈神经网络的流行色预测模型GLMBPNN(Gray Levenberg-Marquardt Back Propagation Neural Network),并利用Levenberg-Marquardt算法提高传统BP(Back Propagation)神经网络模型的学习速率.运用灰色理论学习历史数据的变化规律,对数据进行灰化处理,再对比目标值与BP网络的初始输出值,不断进行逆向反馈修正,训练完毕后通过仿真、白化处理得出流行色预测值.研究表明,GLMBPNN模型预测所得的流行值比灰色模型方法所得值的精度高,且比传统BP神经网络的收敛速度快.

A Fashion Color Forecast Based on Gray Back Propagation Neural Network Model

A new forecasting model—GLMBPNN(Gray Levenberg-Marquardt Back Propagation Neural Network) is proposed,which is based on the gray theory and modified BP(Back Progagation) neural network sped by Levenberg-Marquardt algorithm on learning speed.Firstly,the gray theory is applied to learn changes of historical data and obscure the data by accumulating.Secondly,Levenberg-Marquardt algorithm is utilized to train the BP network with samples made up of processed historical data.Finally,predicted result is attained from simulation and whitening.The research result proves that the GLMBPNN model predict method is more precisely than traditional gray model and possesses a faster convergence speed than the BP network.