An Empirical Study for Adopting Machine Learning Approaches for Gas Pipeline Flow Prediction

Table 4

The details of each performance measure of stacked GBDT models when using the different learning rates.

Learning rate	0.05	0.1	0.15	0.2	0.25
MSE	1.21 × 10^3	1.13 × 10^3	1.34 × 10^3	1.09 × 10^3	8.8 × 10^2
MRE	0.001027	0.000566	0.000717	0.000600	0.000538
	6.423 × 10^−6	5.996 × 10^−6	7.093 × 10^−6	5.777 × 10^−6	4.674 × 10^−6