Gas compressibility factor (z-factor) is an important parameter widely applied in petroleum and chemical engineering. Experimental measurements, equations of state (EOSs) and empirical correlations are the most common sources in z-factor calculations. However, these methods have serious limitations such as being time-consuming as well as those from a computational point of view, like instability, convergence and accuracy. Accurate and fast estimation of this parameter is of interest and a challenging factor in the numerous calculations related to oil and gas processing plants. In this study, a meta-learning algorithm named multi-gene genetic algorithm (MGGP) was applied to predict the sweet gas compressibility factor. To assess the effectiveness of the MGGP model statistical criteria, is applied. The validity of this proposed model was compared with the experimental data. The results showed that the model has successfully predicted the sweet natural gas z-factor, especially at the midrange of operating conditions. However, the MGGP model seems to be inefficient in boundary values of Tpr (i.e. around 1 and 2). In addition, the MGGP model is compared with other z-factor correlations and it is revealed that the implementation of MGGP model lead to a more accurate and reliable estimation of the natural gas compressibility factor.