摘要： 对城市交通流量的特征进行了分析，尝试将核机器方法引入这一领域，对交通流量进行实时预测。综合比较了核机器方法与人工神经网络法的预测效果，同时展示了常规核与复合核的性能对比。实验结果表明，复合核的性能与Sigmoid核和Gaussian核大致相当，稍优于单一的核，为交通实时控制与诱导提供了参考。

关键词: 核机器方法, 交通流量, 人工神经网络, 智能交通系统

Abstract: This paper analyzes features of urban traffic flow, tries to develope kernel machine method in this domain to forecast flow value of future. Artificial neural network and some kinds of kernels are addopted to numerical experiment. Experiment results show that, kernel machine method is better than artificial neural network, and compound kernel functions is better than common single kernel functions. Sigmoid and Gaussian kernel function also show their good performance when contrasted to linear kernel and polynomial kernel function. Kernel machine method finds its applied value in urban traffic real-time control.

Key words: Kernel machine method, Traffic flow, Artificial neural network, Intelligent transportation system

中图分类号: 

• TP301