We provide a brief overview of the modeling process, from Newton’s molecular dynamics to the Boltzmann equation and ultimately the macroscopic Navier-Stokes equations. We then introduce our Mechanism-Data Fusion Method (MDFM) for modeling heat conduction with two dissipative variables, combining the mathematical rigor of physical laws, the adaptability of machine learning, and the solvability of conventional numerical methods. Using the Conservation-Dissipation Formalism, we derive a system of first-order hyperbolic partial differential equations for heat conduction that naturally adheres to the first and second laws of thermodynamics. We train the unknown functions in this system with deep neural networks and propose a novel technique, the Inner-Step Operation, to bridge the gap between the discrete and continuous forms. Extensive numerical experiments show that the model accurately predicts heat conduction across diffusive, hydrodynamic, and ballistic regimes, and outperforms the Guyer-Krumhansl model in terms of accuracy over a wider range of Knudsen numbers.