Analysis and Prediction of Subway Tunnel Surface Subsidence Based on Internet of Things Monitoring and BP Neural Network

Table 5

Accuracy and mean square error of each model.

Network type	The maximum accumulated surface subsidence		The location of the cumulative maximum surface subsidence
	Average relative error (%)	RMS error	Average relative error (%)	RMS error
BP neural network	3.27	3.87	7.96	21.06
GRNN generalized neural network	4.28	10.00	14.38	24.16
RBF neural network	3.27	10.00	17.38	22.70
Elman neural network	7.85	5.29	11.60	18.71