基于ART神经网络的运行轨迹聚类

Trajectory Clustering Based on ART Neural Network

在商用车大数据研究领域,需在相同路况下对车辆的性能、油耗、载重等进行深入研究,那么就需要将相同轨迹的路线取出来,然后去研究这些路线上车辆的行驶状况。文章要在海量的车辆运行数据(具有时间序列)中,对路线进行聚类。首先使用射线法判断GPS点与区域(可视为多边形)的拓扑关系,得到A市到B市之间的GPS数据,然后对GPS轨迹进行网格化处理,将网格划分为0.01度,得到0-1稀疏矩阵,最后建立自适应共振理论(ART)神经网络模型,对99条样本进行聚类,通过调节阈值的大小,得到合适的聚类结果。

In the research field of big data of commercial vehicles,the performance,fuel consumption and load of vehicles should be studied in depth under the same road conditions.Then,the routes with the same trajectory should be taken out and the driving conditions of vehicles on these routes should be studied.In this paper,the route is clustered in the massive vehicle operation data(with time series).Firstly,the topological relationship between GPS points and regions(which can be regarded as polygons)is determined by ray method,and the GPS data between A city and B city is obtained.Then,the GPS track is processed by grid.The mesh is divided into 0.01 degree and 0-1 sparse matrix is obtained.Finally,an adaptive resonance theory(ART)neural network model was established to cluster the 99 samples,and the appropriate clustering results were obtained by adjusting the threshold value.