This paper presents a comparative evaluation of two machine learning models, Gaussian Process Regression (GPR) and Support Vector Regression (SVR), for scattered data approximation. Both models are assessed using key statistical metrics, including the coefficient of determination ( \(R^2\) ), root mean square error (RMSE), and maximum error, with both training and prediction times recorded for analysis. All numerical and graphical results are obtained through MATLAB implementation. The findings reveal that GPR consistently outperforms SVR in terms of approximation accuracy and visualization quality across different functions and data points. Although SVR demonstrates faster training and prediction times, these factors are secondary to the primary objective of achieving accurate approximation. Future work will explore more extensive evaluations involving additional machine learning techniques, larger datasets, and potential applications to real-world data.