基于投影寻踪神经网络模型的短时交通量预测研究

准确有效地预测短时交通流量是实施交通诱导及控制的前提与关键,但由于短时交通流量具有高度复杂性、随机性、非线性和不确定性等特性,导致预测难度高、准确度低、实时性差。基于此,文中综合利用投影寻踪技术和BP神经网络的优点,提出了运用投影寻踪回归模型和BP神经网络技术相结合建立组合模型的预测方法,并编写出模型的算法程序。将该组合模型应用于路段短时交通量的实时预测实例,实验结果证实该组合模型具有较好的预测能力和较强的时效性。

Short-term Traffic Prediction Based on a Combined Projection Pursuit Regression and BP Neural Network Model

Accurately and efficiently predicting the short-term traffic flow is the premise of and key to the successful traffic management and control.However,due to the following characteristics of the short-term traffic flows including complexity,randomness,nonlinearity and uncertainty,prediction of short-term traffic is challenging,with low accuracy and poor real-time performance.Based on this observation,this paper combines the techniques of the projection pursuit regression and BP neural network to establish a combined model,which takes the advantages of both techniques.The solution algorithm of the model is provided.After applying the combined model for real-time short-term traffic,the model is found out to has a better real-time prediction capability.