摘要
道路交叉口作为车辆汇集区域是交通研究的主要对象。利用南京市2010年9月出租汽车GPS数据,通过空间相关分析中Moran′I指数分析了南京市市区空间集聚情况,并利用BP神经网络、支持向量机(SVM)和粒子群优化的支持向量机(PSO-SVM)对空间集聚区域中主干道路形成的交叉口进行了交通流量预测。研究结果表明Moran′I指数能有效找出空间集聚区域,南京市区呈现明显的单中心集聚模式,主要集中在鼓楼区。利用三种模型对市中心交通量进行模拟,结果发现三种模型的相关系数都在0.7以上,但是相较BP神经网络与SVM方法,PSO-SVM的误差最小,误差精度分别提高了约10%和1%,曲线拟合情况与原始数据更加接近。
The road intersection as the vehicles meeting area is the main object of traffic research. The taxi GPS data of Nanjing in September,2010 is used to analyze the spatial agglomeration situation in Nanjing with the Moran′ I index in spatial correlation analysis. The models of BP neural network,support vector machine（SVM）and PSO-SVM are used to forecast the intersection traffic of the trunk roads in spatial agglomeration area. The study results show that the Moran′ I index can effectively find out the spatial agglomeration area,the single center agglomeration mode is presented in Nanjing and located in Gulou district. The traffic of city center was simulated with the above 3 models. The results show that the correlation coefficient of 3 models is above 0.7,the PSO-SVM has the minimum error（the error of PSO-SVM is 10% lower than that of BP neural network and 1%lower than that of SVM method）,and the curve fitting situation is more close to the original data.
出处
《现代电子技术》
北大核心
2016年第17期128-131,共4页
Modern Electronics Technique
基金
国家自然基金:聚落遗址时空演变规律挖掘研究(41071253)
江苏省"六大人才高峰"高层次人才培养对象资助项目(20080249)资助
