We employ the long/shortterm memory (LSTM) recurrent neural network to analyze the impact of various input settings on shortterm traffic flow prediction performance First

we compared the shortterm traffic flow prediction performance for different combinations of traffic flow

speed and occupancy data on the same vehicle detection station (VDS) The results show that the inclusion of occupancy/speed information may help to enhance the performance of the model as awhole In order to introduce spatial information into the model

we further include as inputs traffic variables from the upstream and/or downstream vehicle detector stations and test 16 different input combinations for traffic flow prediction The experimental results show that the inclusion of both upstream and downstream traffic information in the model is very useful for improving the accuracy of shortterm traffic flow prediction