| 摘 要: 为了解决目前市面上的导航软件无法针对用户个性化预测的问题,本文提出了基于边缘计算的道路行程 时间预测的预测思想,并在之后就系统架构和预测过程进行了详细论述。为了实现个性化预测,本文根据速度变化特征 将用户的行为模式分成了三类,并对每一类行为模式构建了对应的数据集。在基于用户行为模式的预测算法方面,本文 对每一类数据集分别应用了ELM模型和LSTM模型,通过对模型的预测表现进行对比确定了最优模型,并将模型及其参 数装载到边缘端。最终,本文通过大量的实验,将该研究所提供的时间预测和百度地图提供的时间预测进行对比,验证了本文研究 内容的可行性和准确性;以实例证明了本文设计并实现的道路行程时间预测原型系统的实用性和有效性。 |
| 关键词: 道路行程时间预测方法;边缘计算;行为模式分类 |
|
中图分类号: TP391
文献标识码: A
|
|
| Road Travel Time Prediction based on Edge Computing |
|
HUANG Xiangming, ZHU Jiahui, ZHU Fukai, GAO Yan
|
( School of Computer Science and Engineering, Northeastern University, Shenyang 110000, China )
20164634@stu.neu.edu.cn; 20164407@stu.neu.edu.cn; 20164448@stu.neu.edu.cn; gaoyan@cse.neu.edu.cn
|
| Abstract: This paper first proposes a framework of road travel time prediction based on edge computation, and then elaborates its system architecture and prediction process. The navigation software in current market does not make personalized prediction for users. In order to achieve personalized prediction, users' behavior patterns will be classi ed into three categories based on characteristics of driving speed changes. For each of these three categories, the corresponding data sets are collected. With regard to prediction algorithm based on users' behavior patterns, the optimal model is determined by comparing the prediction performance of ELM and LSTM models, which are trained with data sets separately. And then, the optimal model and its parameters are loaded to the ends of edge computing. Finally, a large number of experiments are conducted by comparing the time prediction presented in this research with that of Baidu Map so to verify the feasibility and accuracy of this framework. The prototype system of road travel time prediction designed and implemented in this paper is proved to be practical and effective with real cases. |
| Keywords: road travel time prediction method; edge computing; classification of behavior patterns |
