Research Article

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

Table 9

Details of the performance when using different activation functions.

ActivationReLUtan hSigmoidLeaky_ ReLUELU

MSE1.2 × 1087.6 × 1061.2 × 1089.8 × 1037.7 × 106
MRE9.483 × 10−21.210 × 10−29.495 × 10−23.128 × 10−31.858 × 10−2
6.130 × 10−14.035 × 10−26.131 × 10−15.208 × 10−54.081 × 10−2