A new approach for inducing a thermal transport model, of ion heat diffusivity as an example, for magnetically confined high-temperature plasmas has been further pursued after its initial proposal by Yokoyama (2014 Plasma Fusion Res. 9 1302137). It has been based on a statistical approach utilizing the accumulated experimental transport analyses database. Two approaches are described in this paper: (1) placing a priority on reproducing the ion heat diffusivity with higher accuracy for better reproduction of ion temperature profiles, and (2) attempting to acquire a physics interpretation through variable selections and the dependence of the ion heat diffusivity on them. Such progress will foster the study of a practical transport model for real-time control and the provision for guidance to the parameter dependence to be pursued by large-scale cutting-edge simulations.