A Deep Learning Model with Conv-LSTM Networks for Subway Passenger Congestion Delay Prediction

Algorithm 1

The training steps of Conv-LSTM.

	Input Record on the number of inbound passenger flow in the training dataset Record on the number of outbound passenger flow in the training dataset Record on the number of delay rate in the training dataset Record on the number of average congestion delay time in the training dataset Look-back window:
Output Conv-LSTM with learned parameters
Procedure Conv-LSTM training Initialize a null set:  For all defined time slice t do     A training observation is put into
	End for Initialize all the weighted and intercept parameters
	Repeat A batch of samples are randomly selected from to  The parameters are estimated by minimizing the objective function in
	Until Convergence criterion met
End procedure