Short term inbound passenger flow prediction of urban rail transit based on data denoising

In order to improve the accuracy of short-term traffic flow prediction, a combined prediction model based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and Seagull Optimization Algorithm optimized Long-Short Term Memory neural network (SOA-LSTM) was proposed. Firstly, CEEMDAN is used to reduce the interference of data noise, decompose the original passenger flow sequence into multiple characteristic components and residual, and optimize the hyperparameters of the LSTM neural network using SOA. Each feature component is predicted, and the prediction results of each IMFs are summed to obtain the short-term entrance and exit passenger flow prediction results of each station. Extracting four consecutive weeks inbound passenger flow data of Beijing's urban rail transit network in 2019, and validated and compared the proposed short-term passenger flow prediction model. The experimental results demonstrate that compared with the transfer passenger flow prediction results of LSTM, ARIMARBF, LSTM-BP, and SOA-LSTM models, the proposed CEEMDAN-SOA-LSTM model has lower RMSE, MAE, and MAPE error values than other models, with R2 increased by 1.63%, 1.2%, 1.18%, and 0.72%, respectively.