基于加权平均融合模型的出租车订单预测  

Taxi Order Forecasting Based on Weighted Average Fusion Model

作　　者：李清源 朱静[1] 李雨晴 曹海涛[1] 刘彦辰 LI Qingyuan;ZHU Jing;LI Yuqing;CAO Haitao;LIU Yanchen(Xinjiang Agricultural University,Urumiqi 830052)

机构地区：[1]新疆农业大学,乌鲁木齐830052

出　　处：《计算机与数字工程》2024年第12期3585-3589,3609,共6页Computer & Digital Engineering

摘　　要：为提高出租车运营效率,实现对出租车进行合理调度,针对目前单一模型在交通预测问题上精度不高、考虑因素单一的问题,论文在重庆市出租车GPS数据的基础上,加入气象影响因素,研究乘车热点区域的出租车订单数量规律。采用以LSTM、ARIMA、CNN为子模型的加权平均融合模型对热点区域的出租车订单数量进行预测,提高预测精度。结果表明,误差归一化加权平均融合模型不管是相比于其他融合方法还是单一预测模型,都取得了较好的预测结果,更适用于热点区域的出租车需求预测。In order to improve the efficiency of taxi operation and realise reasonable dispatching of taxis,and to address the problems of low accuracy and single consideration of factors in the current single model for traffic prediction problems,this paper adds four meteorological influencing factors on the basis of GPS data of taxis in Chongqing to study the law of taxi orders in the hot-spot areas of ridesharing.A weighted average fusion model with LSTM,ARIMA and CNN as sub-models is used to predict the num-ber of taxi orders in the hotspot areas to improve the prediction accuracy.The results show that the eror normalised weighted aver-age fusion model achieves better prediction results compared to both other fusion methods and single prediction models,and is more suitable for forecasting taxi demand in hotspot areas.

关 键 词：交通预测 融合模型 GPS数据 出租车订单 

分 类 号：U491[交通运输工程—交通运输规划与管理]