基于神经网络的交通量预测技术研究

《基于神经网络的交通量预测技术研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、南京理工大学硕士学位论文基于神经网络的交通量预测技术研究姓名：孟维伟申请学位级别：硕士专业：交通信息工程及控制指导教师：曹从咏20060630AbstractAlongwiththerapiddevelopmentofthetransportationinfrastructureconstructionandintelligenttransportationsystem(ITS)ofourcountry，trafficplanningandtrafficinducingbecametheresearchfocusoftrafficfielddaybyday，whichrealizedwit

2、htheimportantbasisoftheaccuratetrafficvolumeforecasting．Thereforepeoplepaidmoreandmoreattentiontothetrafficvolumeforecasting．ThemainworkofthisthesiswasapplyingtheadvancedartificialintelligencetechnologysuchasartificialneuralnetworkandgeneticarithmeticandSOontothetrafficvolumeforecasting．Itgaveause

3、fulattempttoenrichthemethodsofthetrafficvolumeforecasting．Themaincontentwascomposedwiththefollowingparts：(1)Thedevelopmentaboutartificialneuralnetworkwasintroduced，boththekeystoneandthecorrelationtheoriesaboutartificialneuralnetworkwerestudid，andthekeystoneandarithmeticabouttheBPneuralnetworkweree

4、mphaticallyanalyzed．(2)Theforecastingmethodsforboththelong—termtrafficvolumeandtheshort—termtrafficvolumeweresumlarized，andtheforecastingmechanismandtheshortagesofthemwereanalyzed；thearithmeticflowaboutBPnetworkintrafficvolumeforecastingwaspresented，theidiographicforecastingmethodsforboththelong—·

5、termtrafficvolumeandtheshortl。termtrafficvolumebasedonBPnetworkwereemphaticallystudied．(3)ThemodelingprocessofthetrafficvolumeforecastingmodeIbasedonBPnetworkwasanalyzedatlength，andtheproblemssuchaschoiceaboutthenodenumberofhiddenlayerforBPnetwork，datapretreatment，andSOonwerediscussed；inallusionto

6、theshortagesofBParithmeticsuchaseasytofaliintolocalsmallnessandslowconstringencyspeed，webroughtforwardtheameliorationmeasures．Aftertheabovefoundation，thetrafficvolumeforecastingmodelforthelong-termtrafficvolumeandtheshort—termtrafficvolumebasedonBPnetworkwereconstitutedseparately，theywereappliedby

7、experimentsandachievedperfectresults．H(4)ThegeneticarithmeticandBParithmeticwerecombined．Usingtheadvantageofgeneticarithmeticwhichisgoodatentirelysearching，itwasapplyedtooptimizeinitialconnectionweightsandthresho