The high-precision bus passenger flow prediction is the guarantee of high-quality bus service. The aim of this study is to improve bus passenger flow prediction by comparing and combining different techniques. This study takes Guangzhou bus line 6 as an example to distinguish and analyze characteristics of commuters and non-commuters. Then, several linear and non-linear models are deployed to improve the accuracy of prediction on these two kinds of passenger flow, including ANNS, LSTM and ARIMA model. The prediction result shows that ANNs and LSTM models are separately suitable for predicting commuting flow and non-commuting flow. Based on the prediction performance, this study proposes a hybrid prediction method that combines the ANNs and LSTM model. Several indexes show that the hybrid method is capable of grasping the characteristics of sample passenger flow and can improve the prediction performance in practice.

CICTP