Linear motion control systems are used in a wide range of industrial applications, including machine tools, packaging, textiles, printing, and renewable energy. The main control elements of the linear motion control systems are the motion controller, servo drive, and encoder. The motion controller controls the servo drive coupled with the lead screw or ball-screw assembly in order to obtain the desired motion profile. The motion controllers need to be configured accurately with the parameters of motor revolution, load revolution, and length unit (LU) per load revolution in order to achieve the desired results. The positional errors due to temperature, vibration, backlash, and encoder measurement also affect the accuracy of the linear motion system. In this paper, the proposed experimental data-based modelling of the linear motion system using learning algorithms help the user to configure the motion parameters easily and also reduce the positional errors due to mechanical and measurement error factors. Learning algorithms such as the neural network (NN) and machine learning (ML) algorithms are used to compare and formulate the best model that gives the least root mean squared error (RMSE). The comparison and experimental validation show that the Gaussian process regression (GPR) model outperforms all other models.