面向智能交通系统的空间数据挖掘技术研究-计算机科学与技术专业毕业论文.docxVIP

摘要摘要 摘要 摘要 随着信息技术的发展，智能交通系统和数据挖掘都成为研究热点。智能交通 系统基于GIS(Geographic Information System)技术的应用和发展，拥有了 庞大的空间数据，如何充分利用这些空间数据，从中发掘出隐含其中的知识，进 而依据这些知识进行决策，从而使GIS成为更智能的系统，成为数据挖掘研究的 一个重要内容。交通控制与诱导系统，是ITS研究的热门核心课题，而实现交通 流诱导系统的关键问题是实时准确的交通流量预测。 本文基于深圳市城市交通仿真系统(SUTSS)的空间采集数据支持，结合数 据挖掘技术中的神经网络模型，主要运用了BP神经网络的结构，构造出短时交 通流量的预测模型。其中改进了传统的神经预测完全黑箱操作的模式，引入了交 通流理论的相关概念，对训练样本进行先分类，后学习的方法，提高预测模型的 精准度，一定程度上，避免神经网络收敛速度过慢，陷入局部极小。并通过采用 路段实测交通采集数据，验证所提的算法的有效性。 关键词：智能交通系统，空间数据挖掘，交通流预测，BP神经网络 ABSTRACTWith ABSTRACT With the developing of information technology,Intelligent Transport System and Data Mining have become the hot area in research．Based Oil the development and application of GIS technology，Intelligent Transport System has possessed extremely enormous amount of spatial data．The strategies of making full use of the spatial data and exposing the potential knowledge behind the data for supporting the decision-making，which can make GIS more intelligent，have been a significant commitment of data mining research． Based on the spatial data from Shenzhen Urban Transport Simulation System， adopting Neural Network theory of Data Mining，especially Error Back Propagation Neural Network，the model for short-term traffic flow forecasting has been structured in this research．The model improves the traditional Neural Network algorithm which is totally established in a black box way,by introducing the conception of traffic flow theory．Before training the neural cells，the sample space will be classified into different set．In this way,the BP Neural Network will produce more exact result for traffic flow forecasting，and to some extent，the model also Can prevent the slow convergence and local minimum which tend to occur in the BP Neural Network algorithm．At the end，Through the traffic data collected from the real road in SUTSS， the efficiency of the proposed algrithom has been validated． Key Words：Intelligent Transport System，Spatial Data Mining，Traffic Flow Forecasting，BP Neural Network II 学位论文版权使用授权书本人完全了解同济大学