基于个性诊断的路段缺失速度值估计

Estimating missing road speeds with personality diagnosis algorithm

针对基于GPS的浮动车技术因无法做到对路网的时空间全覆盖导致部分路段实时交通状态缺失问题,提出基于热门路段个性化诊断(personality diagnosis base on popular road,PDPR)模型对各路段上的缺失速度值进行估计。使用K均值算法对所有原始数据作离散化处理,根据数据覆盖率对路段进行分类;以高覆盖率路段的速度数据为辅助,使用个性诊断算法(personality diagnosis,PD)对低覆盖率路段进行缺失速度估计,把估计值映射到连续型空间。实验结果表明,PDPR模型估计误差比PPCA(probabilistic principal component analysis)算法低32.84%,比滑动平均法低5.70%。

The missing road speeds estimation problem, which arises for reason that the amount of floating vehicles-the dominating means used in real-time traffic speed collection is far less than that of the roads. Personality diagnosis based on popular road （PDPR） is proposed to solve this problem. PDPR is summarized in three stages： First, the k-means algorithm is used to cluster all the speed data into a few non-overlapped groups, for data within the same cluster, replace them with the group mean; Secondly, partition the road set into two subsets with the data coverage on the roads, as the result, one of two sets consists of roads with relatively high data coverage, while the other one are roads of low coverage. At last, the personality diagnosis algorithm is used to make missing estimations for all roads with auxiliary from the subset of roads with high coverage. Experimental result studies show that the estimated error of PDPR is 32. 84% lower than PPCA （probabilistic principal component analysis）, and 5.7% lower than the moving average model.