摘要
为准确地对交通流进行短时预测,提出了一种新的基于数据挖掘技术的预测模型。该模型综合了改进遗传算法、粗糙集理论和小波神经网络三种数据挖掘技术。建模阶段分为离散化、属性约简和训练三个步骤。在离散化的过程中,采用了改进的遗传算法,保证了系统分类能力,且使断点数目最少;属性约简中采用了粗糙集理论,选择对交通流预测密切相关的属性,加速了小波神经网络的预测速度并使其结构简化;训练中利用了小波变换非线性特性,采用经过属性约简后的数据对小波神经网络进行训练,从而获得短时预测模型。为验证模型的有效性,进行了对比测试,分析结果证实了该预测模型比传统方法具有更高的精度和速度,为交通流的准确实时预测提供了一种新的思路。
To forecast short-term traffic flow accurately, the forecasting model was proposed based on data mining technology. The model improved three data mining technology as rough set theory, genetic algorithm and wavelet neural networks. There were three steps in modeling phase, as discretion, attribute reduction and training. Firstly, the improved genetic algorithm was used to discrete continuous attributes which remained the discernable ability of decision system with least numbers of broken points. Then, decretive data were reduced by using the rough set theory so as to improve forecasting speed and simplify the structure of networks. Finally, the data reduced were input to nonlinear wavelet neural networks. By comparative testing, higher precision and speed were achieved by using the data mining model, which provided a new idea for short-term traffic flow forecasting.
出处
《计算机集成制造系统》
EI
CSCD
北大核心
2008年第4期690-695,共6页
Computer Integrated Manufacturing Systems
基金
国家科技支撑计划资助项目(2006BAF01A46)
上海市社会发展重大专项资助项目(06DZ12001)
上海市基础研究重点资助项目(06JC14066)
上海市科技发展基金重点资助项目(061612058)
上海市登山行动计划资助项目(061111006)~~
关键词
数据挖掘
预测模型
粗糙集
遗传算法
小波神经网络
data mining
forecasting model
rough set
genetic algorithm
wavelet neural networks
