多层混沌神经网络及其在交通量预测中的应用

研究多层混沌神经网络及其在交通量预测中的应用问题。以BP网络和混沌理论为基础,提出了一种在隐层中包含混沌神经元的多层混沌神经网络。XOR问题实验得出:该混沌神经网络能有效地强化网络的非线性和学习效率。鉴于城市交通流具有明显的混沌特性,将该混沌神经网络应用于城市交通流的预测。对广东江门市某路口交通量的预测结果显示出:采用该混沌神经网络,预测误差一般可以控制在10%以下(或左右)。该网络还可以应用于其他混沌系统的预测和控制。

Multi-layer Neural Network Involving Chaos Neurons and Its Application to Traffic-flow Prediction

To investigate multi-layer chaos neural networks (MLCNN) and its application in traffic-flow forecasting,a multi-layer chaos network involving chaos neurons in hidden layer was developed based on BP network and chaos theory. The learning ability of the MLCNN was evaluated by benchmark experiment with XOR problem,and the experimental results show the feasibility of the chaos neurons to reinforce the learning ability and non-linear of the network. Urban traffic system could be regard as a chaotic system,so the MLCNN was used in prediction to traffic flow. The test result to certain intersection in Jiangmen city shows that the prediction error could be controlled below (or around) 10%. The MLCNN could also be used in prediction and control to the other chaotic systems.