基于大规模数据的公交到站时间预测方法比较-《计算机技术与发展》

文章信息/Info

Title:: Empirical Comparison among Classic Bus Arriving Time Prediction Methods Based on a Large-scale Dataset

作者:: 庞俊彪¹; 胡安静¹; 黄晶¹; 杜勇²; 于海涛²; 1. 北京工业大学信息学部北京多媒体与智能软件技术重点实验室,北京 100124;2. 北京市交通信息中心,北京 100161

Author(s):: PANG Jun-biao¹; HU An-jing¹; HUANG Jing¹; DU Yong²; YU Hai-tao²; 1. Beijing Key Laboratory of Multimedia and Intelligent Software Technology,Department of Information, Beijing University of Technology,Beijing 100124,China;2. Beijing Transportation Information Center,Beijing 100161,China

Keywords:: bus arrival time prediction; performance comparison; algorithm evaluation; GPS data

摘要:: 公交到站时间预测作为提高公共交通运输服务水平的重要措施,能够鼓励用户使用公共交通出行,方便调度部门进行合理调度。通过研究现有文献,发现虽然已经提出了很多不同原理的公交到站时间预测方法,但由于各个文献中使用的数据集合、测试规模不同,所以在现有方法之间无法进行有效的比较,从而无法发现公交到站时间预测的基本问题。为了提供可靠准确的数据基础,实现在统一的数据集上公平地比较目前现有的方法,建立了北京市公交到站数据集。该公交到站数据集是目前为止最大的公交运营数据集,其中包含了各种复杂的路况和可能的情况。在北京市公交到站数据集上,通过选择典型到站预测方法,进行实验比较和结果分析,定位出公交到站时间预测的本质问题。

Abstract:: As an important measure to improve the level of public transport service,the prediction of bus arriving time can encourage users to choose public transport and facilitate the scheduling department to make reasonable scheduling. Via the investigation of various existing methods for bus arriving time prediction,however resulting from using diverse datasets with different scale,the comparisons among them are not effective,so the intrinsic of bus arriving time prediction model cannot be found. Due to above shortcomings,the most comprehensive Beijing bus transportation dataset which includes sundry road conditions and providing reliable accurate platform to compare the algorithm is built. Based on the Beijing bus transportation dataset,through the analysis and comparison of existing prediction algorithms,as a result,the essential problem of bus arriving time prediction model is found.