首页|地铁短时客流预测改进LSTM方法

地铁短时客流预测改进LSTM方法

扫码查看
短时客流预测可为轨道交通运营部门规划调度提供参考,其中短时客流预测的精准性尤为重要,为进一步提高城市轨道站点短时客流预测精准性,提出一种结合集合经验模式分解算法和贝叶斯优化算法的改进 LSTM方法.先使用集合经验模式分解算法(EEMD)对地铁站点的客流数据进行分解,以减少数据噪声干扰;再通过贝叶斯优化算法(BOA)对长短时记忆神经网络(LSTM)的超参数进行优化,从而提高模型的参数精确性.采用真实的客流数据验证结果表明:相较于单一 LSTM以及单层组合模型,双重叠加后的 EEMD-BOA-LSTM组合模型预测结果平均绝对误差降低 21.8%~44.8%,均方根误差降低 16.9%~47.4%,对短时客流的预测结果误差改善显著.
An improved LSTM method for short-term passenger flow prediction in subways
In order to further improve the accuracy of short-term passenger flow prediction of urban rail stations,an improved LSTM method combining ensemble empirical pattern decomposition algorithm and Bayesian optimization algorithm was proposed.Firstly,the EEMD is used to decompose the passenger flow data of subway stations to reduce the interference of data noise.Then,the BOA is used to optimize the hyper parameters of the LSTM,so as to promote the parameter accuracy of the model.Compared with the single LSTM and single-layer combination model,the testing results of prediction show that the Mean Absolute Error(MAE)and Root Mean Square Error(RMSE)of the two-layer combined model named EEMD-BOA-LSTM,are respectively reduced by 21.8%~44.8%,16.9%~47.4%,the error of the prediction results for short-term passenger flow has been significantly improved.

rail transitshort-term passenger flow predictionlong short-term memory neural networksensemble empirical mode decompositionBayesian optimization algorithm

亓晓雨、傅成红

展开 >

福建理工大学 交通运输学院,福州 350118

福建理工大学 交通运输研究所,福州 350118

轨道交通 短时客流预测 长短时记忆神经网络 集合经验模态分解 贝叶斯优化算法

福建省自然科学基金项目福建理工大学科研项目

2020J05194GY-H-21153

2024

交通科技与经济
黑龙江工程学院

交通科技与经济

影响因子:0.862
ISSN:1008-5696
年,卷(期):2024.26(2)
  • 26