摘要
为改善浮动车全球定位系统(global positioning system,GPS)数据因采集过程中受到干扰造成数据缺失问题,通过分析法研究了浮动车GPS数据与交通流状态和道路线形之间的关联性,提出一种基于优化随机森林算法的浮动车GPS数据插补模型,该模型针对随机森林算法插补过程中,因自身的随机性而引起插补结果具有波动性问题,在结果输出部分引入权重因子,通过线性优化算法,调节权重因子大小使输出结果波动性降低的同时满足道路线形特征。实验对6名志愿者21 d的出行轨迹数据进行插补。结果表明:所构建的模型平均误差为12.3 m,相较于随机森林模型、决策树模型和线性回归模型分别减少14.9、24.3、239.3 m,可见采用优化随机森林算法建立的插补模型有效提升了浮动车GPS数据插补精度,为交通状态分析、地图匹配等应用提供数据基础。
In order to improve the problem of data missing due to the interference of the floating car global positioning system(GPS)data during the collection process,the correlation between the floating car GPS data and the traffic flow state and road alignment was studied through the analysis method,the floating car based on the optimized random forest algorithm was proposed.In the random forest algorithm interpolation process,the interpolation result had volatility problems due to its own randomness,and a weight factor was introduced in the result output part.Through the linear optimization algorithm,the weight factor was adjusted to reduce the volatility of the output result while satisfying the road alignment characteristics.The experiment interpolated the travel trajectory data of 6 volunteers for 21 days.The results show that the average error of the model constructed is 12.3 m,which decreased by 14.9,24.3,239.3 m,respectively compared with the random forest model,decision tree model and linear regression model.It is concluded that the interpolation model established by the optimized random forest algorithm can improve the accuracy of the floating car GPS data interpolation,and provides a data basis for applications such as traffic state analysis and map matching.
作者
吕勤学
郭杜杜
李心
赵亮
LÜQin-xue;GUO Du-du;LI Xin;ZHAO Liang(School ofmechanical engineering, Xinjiang University, Urumqi 830046, China)
出处
《科学技术与工程》
北大核心
2022年第4期1656-1661,共6页
Science Technology and Engineering
基金
新疆维吾尔自治区自然科学基金(2019D01C043)。
