Feasibility of activation energy prediction of gas-phase reactions via machine learning.