主办:陕西省汽车工程学会
ISSN 1671-7988  CN 61-1394/TH
创刊:1976年

汽车实用技术 ›› 2023, Vol. 48 ›› Issue (16): 188-193.DOI: 10.16638/j.cnki.1671-7988.2023.016.039

• 标准·法规·管理 • 上一篇    

基于深度学习的网约车需求预测研究

全煜坤   

  1. 长安大学 汽车学院
  • 出版日期:2023-08-30 发布日期:2023-08-30
  • 通讯作者: 全煜坤
  • 作者简介:全煜坤(2000-),男,硕士研究生,研究方向为深度学习,双挂汽车列车稳定性,E-mail:qyk000709@163.com。

Research on Demand Forecast of Online Car-hailing Based on Deep Learning

QUAN Yukun   

  1. School of Automobile, Chang'an University
  • Online:2023-08-30 Published:2023-08-30
  • Contact: QUAN Yukun

摘要: 网约车需求预测是一个典型的时间序列预测任务,准确的网约车需求预测能够辅助网 约车平台合理地派单和规划路径,从而降低网约车的空驶率,具有重要的研究意义。文章利 用长短时记忆模型(LSTM)及门控循环单元(GRU)进行网约车需求预测,对比了同一地 区休息日和工作日,一周和一个月内的网约车需求及其变化,构建基于 LSTM 和 GRU 的需 求预测模型,使用历史数据对未来需求进行预测,使用 Geohash 代码对西安市进行区域划分, 对数据和划分的网格进行匹配得到汇总数据,采用线性模型进行对照试验,结果表明,LSTM 和 GRU 在网约车需求预测中的表现优于线性模型,二者相比 LSTM 预测精度更高。

关键词: 网约车需求预测;时间相关性;深度学习;长短期记忆神经网络

Abstract: The demand forecasting of online car-hailing is a typical time series forecasting task. The accurate demand forecasting can help the platform to arrange the order and plan the route reasonably, so as to reduce the empty driving rate of ride-hailing. It has important research significance. In this paper, long short-term model (LSTM) and gated recurrent unit (GRU) are used to forecast the demand of online car-hailing, demand for online car-hailing within a week and a month and its changes, building demand forecasting models based on LSTM and GRU, using historical data to predict future demand, and using Geohash code to map Xi'an by region, the data are matched with the grid to get the summary data, and the linear model is used for the control test. The results show that the performance of LSTM and GRU is better than that of the linear model in the demand forecast of online car-hailing, the prediction accuracy of LSTM is higher than that of LSTM.

Key words: Demand forecast of online car-hailing; Temporal correlation; Deep learning; Long short- term memory neural network