基于VMD-ISSA-BiLSTM的高速铁路短时客流预测OA

To aim at the problems of dynamic fluctuations and non-stationary characteristics of high-speed railway passenger flow,as well as the insufficient robustness of traditional prediction models in complex environments,a short-term passenger flow prediction model that integrates multi-scale data features was proposed.The model first adopted the variational mode decomposition algorithm to decompose the original passenger flow time series data into multiple modal functions with different frequencies and complexities.Next,an improved sparrow search algorithm was employed to enhance important hyperparameters in the bidirectional long short-term memory network,which enhancing the model's ability to capture temporal dependencies.A fully connected layer was then used to integrate multi-dimensional forecast results.The combined prediction model was constructed.Finally,passenger ticket data from the Guangzhou-Zhuhai Intercity Railway was used as a case study.The model was compared with seven classic prediction models at three different time granularities of 15 minutes,30 minutes,and 60 minutes.The calculation results show that the proposed model can exhibit good and stable forecasting performance at different time granularities,with its advantages being more pronounced at shorter time granularities.At the 15-minute time granularity,compared to the Convolutional Long Short-Term Memory(ConvLSTM),Gated Recurrent Unit(GRU),and Autoregressive Integrated Moving Average(ARIMA)models,the coefficient of determination can increase by 5.18%,36.49%,and 54.74%,respectively,for each comparison.In the ablation experiments,after removing the variational mode decomposition algorithm branch,the model's mean absolute error(fmae),root mean square error(frmse),and mean absolute percentage error(fmape)metrics can increase by 39.035,36.474,and 2.699,respectively.After removing the parameter dynamic optimization branch,the model's fmae,frmse,and fmape evaluation metrics can increase by 87.992,99.708,and 11.315,respectively.The proposed model in this paper can effectively capture the multi-scale features of passenger flow.The method can provide reliable data support for high-speed rail transportation organization and management decision-making,contributing to the enhancement of transport service quality and passenger satisfaction.