This paper begins by examining two simple classical nonlinear systems—the double pendulum system and the Rssler system. Through the use of the SINDyPI algorithm for identifying their dynamical equations, it demonstrates the efficiency and accuracy of machine learning methods in handling nonlinear systems and highlights the merits of this algorithm in datadriven approaches. Additionally, taking this as a basis, the study explores and discusses effective approaches to incorporating machine learning into the computational physics course. The findings provide valuable references for the practical implementation of teaching in computational physics.