混合神经网络挖掘模型在交通流量预测中的研究  

Study on Hybrid Neural Network Data Mining Model for Traffic Flow Prediction

作　　者：杨凡 严忠贞[2] 

机构地区：[1]广州女子职业技术学院,广州511450 [2]武汉理工大学智能交通研究中心,武汉430063

出　　处：《小型微型计算机系统》2012年第9期1978-1981,共4页Journal of Chinese Computer Systems

基　　金：国家"九七三"重点基础研究发展计划项目(2005CB724205)资助;交通部行业攻关项目(2009-353-460-640)资助

摘　　要：交通流量预测是交通控制与交通诱导的关键技术,然而对于实现准确流量预测的可靠知识隐藏在大量的交通数据之中,需要对海量数据进行挖掘以发现潜在流量变化规律.传统的交通流量预测主要依靠专家经验对数据进行类别标记,其预测结果受到专家知识限制的影响较大.为了减轻人为因素的影响,提出一种混合智能数据挖掘的交通流量预测模型.首先利用自组织神经网络(SOM)的无监督学习方式实现海量数据类型特性的自动标识,降低对专家经验的依赖度;其次采用改进遗传算法(GA)优化模糊神经网络(FNN),对标识数据进行学习,建立交通流量预测模型.通过对智能交通系统(ITS)的实际数据进行分析,结果表明本文所提出的数据挖掘方法准确有效,预测精度达到95%,比不使用遗传算法优化提高了近8%.Potential knowledge useful for traffic management optimization is hidden in a huge amount of data. Previous works use the prior data pattern labels to train and attain the intelligent data mining models. The performance of the models suffers from the experts＇ experience. To relieve the impact of the human factor, a new hybrid intelligent data mining model is proposed in this work based on self-organizing map （SOM） and fuzzy neural network （FNN）. The SOM was firstly used to capture the clustering information of the database through an unsupervised manner. Then the identified samples were treated as input to train the FNN. In addition, the im- proved genetic algorithm （GA） was employed to tune the FNN parameters and hence the satisfactory intelligent model was obtained. The Intelligent Transportation Systems （ITS） were applied to the validation of the proposed mining model. The analysis results show that the proposed method can extract the underlying rules of the testing data and can predict the future traffic state with the accuracy beyond 95%. which is increased by 8% agalnst traditional method.

关 键 词：智能交通系统 交通流量预测 数据挖掘 人工智能 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]