面向城市轨道交通运营管理的短时进站客流预测研究

Research on short-term inbound passenger flow prediction for urban rail transit operation management

在使用数据挖掘进行数据预处理的基础上,通过卷积神经网络(CNN)、深度残差神经网络(ResNet)和双向长短时记忆网络(BiLSTM)对客流数据进行分析并生成客流预测数据。结果显示,进行100次训练后的训练损失值为10;进行客流量预测时所得预测结果在0人次到2000人次范围内和真实值基本拟合;评价绝对百分比误差仅为12.77737%。说明研究方法具有较高的训练效率,能较准确地进行面向城市轨道交通运营管理的短时进站客流预测。

On the basis of data preprocessing using data mining,the study analyzes passenger flow data and generates passenger flow forecast data through Convolution Neural Network(CNN),Deep residual network(ResNet)and Bidirectional Long Short-Term Memory Network(BiLSTM).The results show that the training loss value of the research method after 100 times of training is 10.The prediction results obtained during the passenger flow prediction range from 0 to 2000 people are basically fitted with the real value;the absolute percentage error of the evaluation is only 12.77737%.It shows that the research method has high training efficiency and can more accurately predict short-term inbound passenger flow for urban rail transit operation management.