基于改进LSTM的城市轨道短时流量预测研究
Research on Short-term Flow Prediction of Urban Rail Based on Improved LSTM
魏化永 1李建华2
作者信息
- 1. 安徽交通职业技术学院 城市轨道交通与信息工程系,安徽 合肥 230051
- 2. 合肥工业大学 计算机与信息学院,安徽 合肥 230022
- 折叠
摘要
针对 LSTM应用于城市轨道短时流量预测存在的模型参数确定困难、对预测精度影响大的问题,采用改进的 BA 算法对模型参数进行优化.对传统 BA 算法,采用自适应策略来动态调整脉冲频率和蝙蝠速率,同时蝙蝠位置更新模型中引入随机扰动项,提高了 BA 的优化性能.采用 BA 对 LSTM参数进行优化,提出了基于改进 LSTM的城市轨道短时流量预测模型.将提出的模型应用于郑州地铁 1 号线,通过和 BP 神经网络预测模型、LSTM 预测模型的对比,验证了所提出的改进 LSTM预测模型具有更高的预测精度.
Abstract
Aiming at the difficulties in determining the model parameters of LSTM applied to urban rail short-term flow prediction,which has great influence on prediction accuracy,the im-proved BA algorithm is used to optimize the model parameters.For the traditional BA algorithm,the adaptive strategy is used to dynamically adjust the pulse frequency and bat rate,and the ran-dom disturbance term is introduced into the bat position update model,which improves the opti-mization performance of BA.Using BA to optimize LSTM parameters,a short-term urban rail flow prediction model based on improved LSTM is proposed.The proposed model is applied to Zhengzhou Metro Line 1,and compared with BP neural network prediction model and LSTM pre-diction model;the improved LSTM prediction model proves to have higher prediction accuracy.
关键词
改进蝙蝠算法/长短时记忆网络/城市轨道交通/短时客流量预测Key words
improved bat algorithm/long short-term memory network/urban rail transit/short-time passenger flow prediction引用本文复制引用
基金项目
安徽省教育厅质量工程项目(2021gkszgg018)
安徽省高校自然科学研究项目(2023AH040384)
出版年
2024
NETL
NSTL
万方数据