【摘要】 首次将神经网络集成技术引入交通量预测。神经网络集成通过训练多个神经网络并将各网络输出进行合成,可显著提高学习系统的泛化能力。在Boosting和Bagging集成方法的基础上,提出基于分治策略的神经网络集成方法,并且讨论了网络权重分配算法。使用上述三种神经网络集成预测模型,对苏州某交叉口实时交通量进行预测,预测结果比较理想,优于单一神经网络预测方法。实验表明,神经网络集成用于交通量预测是有效可行的。更多还原

【Abstract】 Neural Network ensemble is firstly applied to forecast traffic flow,which can improve remarkably the generalization ability of learning systems through training several neural networks and then combining their results.Based on Boosting and Bagging,the method of neural network ensemble with the strategy of divide and conquer is proposed,and the assignment algorithm of sub-neural networks weight coefficients is also discussed.The above-mentioned three models are employed to forecast traffic flow of an intersection in Suzhou city with favor resul.The experiments show that neural network ensemble method is better than simplex neural network,and traffic flow forecasting based on neural network ensemble is valid and feasible.更多还原