The invention discloses a subway line network passenger flow prediction method. 2, data preprocessing: analyzing long-term trend change, seasonal change, weekly cyclic change and abnormal change characteristics of predicted input data, and preprocessing the predicted input data; and step 3, normal daily passenger flow prediction: (1) according to the preprocessed time sequence data, establishing an ARIMA passenger flow prediction model, performing parameter calibration and model optimization inspection, outputting a passenger flow prediction result and performing index evaluation; (2) according to the preprocessed time sequence data, establishing an LSTM artificial neural network passenger flow prediction model, carrying out batch training, setting a learning rate lr and training times epoch, outputting a fitting result and carrying out reverse normalization, carrying out multiple rounds of training, calculating an evaluation index and optimizing model parameters, carrying out passenger flow prediction and outputting a prediction result; and 4, holiday and festival passenger flow prediction. The method is short in time consumption and stable in passenger flow prediction level.
本发明公开了一种地铁线网客流预测方法,包括以下步骤;步骤1、客流数据采集:步骤2、数据预处理:对预测输入数据的长期趋势变化、季节性变化、周循环变化和非常态变化特征进行分析,对预测输入数据进行预处理;步骤3、平常日客流预测:(1)根据预处理后的时间序列数据,建立ARIMA客流预测模型,进行参数标定和模型优化检验,输出客流预测结果并进行指标评价;(2)根据预处理后的时间序列数据,建立LSTM人工神经网络客流预测模型,进行分批训练,设置学习速率lr和训练次数epoch,输出拟合结果并反归一化,进行多轮训练,计算评价指标并优化模型参数,进行客流预测并输出预测结果;步骤4、节假日客流预测:本发明耗时较短,客流预测水平稳定。
Subway line network passenger flow prediction method
一种地铁线网客流预测方法
2023-04-11
Patent
Electronic Resource
Chinese
Subway section passenger flow prediction method based on LSTM network
European Patent Office | 2022
|Subway transfer passenger flow prediction method, device and equipment
European Patent Office | 2023
|Hybrid models of subway passenger flow prediction based on convolutional neural network
DOAJ | 2023
|Hybrid models of subway passenger flow prediction based on convolutional neural network
Wiley | 2023
|Short-term subway station entering passenger flow volume prediction method
European Patent Office | 2023
|