基于模糊神经网络卡车路段行程时间实时动态预测

Real-time dynamic forecasts of truck link travel time based on fuzzy neural network

提出了一种基于模糊神经网络卡车路段行程时间实时预测模型,阐述了自适应神经网络模糊系统(Adaptive Network-based Fuzzy Inference System,ANFIS)网络原理和方法对行程时间预测的可行性和可靠性,采用最小二乘法和误差反传算法结合的混合学习算法,减少了搜索空间的维数,而采用的减法聚类方法减少了模糊推理规则.混合学习算法和减法聚类方法的应用提高了网络参数的辨识和收敛速度.实例仿真论证了该模型预测速度更快、准确性更高,实时性好,获得了比单纯使用神经网络或模糊理论更精确的预测结果.

Put forward a real-time dynamic truck link travel times forecasting model based on fuzzy neural network, discussed the theory and method of adaptive network-based fuzzy inference system （ANFIS） network, and the feasibility and reliability of forecasting the travel time. The hybrid learning algorithm which combines error back propagation algorithm with least-square estimation was used. It makes the dimension of searching space be reduced. The fuzzy inference rule is decreased by using subtraction clustering method. This hybrid learning algorithm and subtraction clustering method greatly raise the speed of parameter identification and convergence. The simulation result shows that the ANFIS network model is more accurate than pure neural network or pure fuzzy theory application, its speed becomes more faster, its accuracy becomes more higher and better real-time.