Dew point pressure, at which the first condensate liquid comes out of solution in gas condensate reservoir, is a vital parameter for fluid characterization, field development, reservoir management and facility construction. Fast and accurate measurement of dew point pressure is always a challenge. Laboratory measurement can give accurate dew point pressure, but are expensive and time consuming. Equation of state is an alternative way, but can not converge in light oil and gas condensate reservoirs. Different empirical correlations have been built up between reservoir properties, fluid composition and dew point pressure. However, those correlations do not accurately reflect complex, non-linear relationships between them. With the development and improvement of artificial neural networks, different neural networks; such as multilayer perceptron neural network, radial basis function neural network, and gene expression programming can be used to describe complex relationships. Recently, one popular machine learning algorithm-(support vector machine) attracts attention due to its strong generalization ability. In this paper, we introduce a mixed kernel function based support vector machine (MKF-SVM), which has both strong interpolation and extrapolation abilities. This support vector machine model was trained and tested using 564 measurements of dew point pressure. The performance of this model is compared against four well known empirical correlations for dew point pressure calculation. The result, high R-2 = 0.9150, low root mean square error RMSE = 476.392 and low average absolute percent relative error (AAPE = 7.01%) indicates good performance of mixed kernel function based support vector machine (MKF-SVM).