交通量的灰色神经网络预测方法

结合灰色系统思想与神经网络构成灰色神经网络,根据目前灰色模型与神经网络结合的方法,提出并联型、串联型和嵌入型3种预测模型的结构.并联型灰色神经网络首先采用灰色模型、神经网络分别进行预测,而后对预测结果加以组合作为实际预测值;串联型对多个灰色预测的结果使用神经网络进行组合;嵌入型在神经网络的输入端、输出端分别增加一个灰化层和白化层而构成.对并联型灰色神经网络给出一种根据预测模型的有效度确定加权系数的方法.将上述3种灰色神经网络模型用于对京石高速公路断面机动车实时交通量进行预测,模型精度和预测结果比较理想,优于单一预测模型.实验表明:灰色神经网络可提高预测精度,用于交通量预测方法是有效可行的.

Grey neural network forecasting for traffic flow

Grey neural network (GNN)combines grey system with neural network. There are three kinds of forecasting model structure: parallel grey neural network (PGNN), series grey neural network (SGNN) and inlaid grey neural network (IGNN). PGNN uses grey model and neural network to predict separately, then combines the predicting results; SGNN employs grey model to predict, then uses neural network to combine the predicting results; IGNN is built by adding a grey layer before neural input layer and a white layer after neural output layer. According to the effectiveness indicator of the forecasting model a method for calculating weight coefficients in grey neural network model is given. The above three GNN models have been employed to forecast a real vehicle traffic volume in Jingshi highway with satisfied precision. The experiments show that the GNN models overmatch the single GM model or neural network, therefore traffic volume forecasting based on GNN is feasible.