The accurate calculation of transport coefficients in high-speed flow modeling of reacting air is essential in the analysis of heat transfer processes in various problems of gas dynamics, for example, in design of thermal protection of descent vehicles. The most accurate from the physical point of view within the continuum approximation is the state-to-state approach. However, it has a significant computational complexity. One of the most effective and accessible method of applying such models is a machine learning regression. We study the application of this approach to the problem of modeling viscosity and thermal conductivity coefficients. Further prospects of machine methods in modeling of high-speed flows are considered.