Abstract: Newton-Raphson iteration is used in semi-classical model of carbon nanotube field effect transistors(CNTFET) based on ballistic transport for self-consistent potential.In each iteration re-integration of state density is requred.Support vector regression (SVR) is applied for the relationship of self-consistent potential and carrier density.Numerical integrations are avoided.And gradient descent algorithm (GDA) is used to get self-consistent potential.It shows that compared with Newton-Raphson iteration,the proposed method reduces computation effectively while maintains high accuracy.It lays theoretical foundation for designing and applying CNTFET devices in large scale integrated circuits.