An EEMD-EWT-LSTM-based short-term prediction approach for inbound metro ridership.

There is no doubt that accurate short-term inbound metro ridership forecasting has always been a significant part in terms of intelligent transportation. Based on the excellent performance of deep learning models in dealing with processing time series data, a large number of researchers have adopted deep learning methods for short-term inbound metro ridership prediction. However, considering the non-linear and cyclical nature of inbound metro ridership, the prediction accuracy using only deep learning models is limited. To solve this problem, a hybrid prediction model combining Ensemble Empirical Mode Decomposition (EEMD), subsequences recombination, empirical wavelet transform (EWT) and long short-term memory (LSTM) network for inbound metro ridership which can greatly reduce the complexity of the original inbound metro ridership is proposed in this article. To prove the superiority of the proposed EEMD-EWT-LSTM-based model, we compare it with recent main forecasting models on three real datasets from the 2019 TIANCHI competition. The experimental results demonstrate that the reduction in WMAPE of the hybrid model almost reach to 10 percent compared to the non-hybrid models. [ABSTRACT FROM AUTHOR]