基于BOA-LSTM模型的地铁站客流预测

杨鑫宇^1,2，陈队永^2,3

1.河北科技学院汽车工程学院，河北 唐山  063200；

2.石家庄铁道大学河北省交通安全与控制重点实验室，河北 石家庄  050043；

3.石家庄铁道大学交通运输学院，河北 石家庄  050043

摘要：针对地铁站客流预测方法单一、预测准确度较低等问题，基于贝叶斯优化算法（Bayesian optimization algorithm，BOA）的全局寻优能力，寻找长短时记忆（long short time memory，LSTM）神经网络的最优超参数，提出BOA-LSTM客流预测模型。以石家庄地铁1号线北国商城站为例，根据该站客流特征分别采用自回归移动平均(autoregressive integrated moving average，ARIMA）模型、LSTM神经网络和BOA-LSTM模型预测2021年7、8月工作日与自然日的进、出站客流。结果表明：BOA-LSTM模型的客流量预测结果与实际客流量的平均绝对百分比误差、平均绝对误差、均方误差、均方根误差均最小，客流预测准确度较高；BOA-LSTM客流预测模型的适用期为1~2个月，在地铁站短时客流预测中具有较强的实用性。

关键词：地铁站；客流预测；BOA；LSTM神经网络；准确度

Passenger flow prediction of subway stations based on

BOA-LSTM model

YANG Xinyu^1,2, CHEN Duiyong^2,3

1. College of Automotive Engineering, Hebei University of Science and Technology, Tangshan 063200, China;

2. Key Laboratory of Traffic Safety and Control of Heibei Province, Shijiazhuang Tiedao University, Shijiazhuang 050043, China;

3. School of Traffic and Transportation, Shijiazhuang Tiedao University, Shijiazhuang 050043, China

Abstract:In response to the issues of single method and low prediction accuracy in subway station passenger flow prediction, a Bayesian optimization algorithm (BOA)-LSTM passenger flow prediction model is proposed based on the global optimization ability of BOA and the long short term memory (LSTM) neural network. Taking the Beiguoshangcheng Station of Shijiazhuang Subway Line 1 as an example, the autoregressive integrated moving average (ARIMA) model, LSTM neural network, and BOA-LSTM model are used to predict the inbound and outbound passenger flows on working days and non-working days in July and August 2021 based on the characteristics of passenger flow at the station. The results show that the BOA-LSTM model has the smallest average absolute percentage error, average absolute error, mean square error, and root mean square error compared to the actual passenger flow, indicating higher prediction accuracy. The BOA-LSTM passenger flow prediction model is applicable from 1 to 2 months and has strong practicality in short-term passenger flow prediction at subway stations.

Keywords:subway station; passenger flow prediction; BOA; LSTM neural network; accuracy