Fast and precise prediction of the dew point pressure in condensate gas reservoirs was of great significance for making up gas reservoir development plans. Traditional methods for predicting the dew point pressure in condensate gas reservoirs include the experimental method, the equation of state method and the empirical formula method, but the traditional dew point pressure prediction method has some problems, such as the experimental method has high experimental test cost, time-consuming and labor-intensive, the empirical formula method has poor universality and weak generalization, the equation of state method needs to fit the binary coefficients of each component, and there are certain problems in the convergence of the solution. Therefore, consider employing machine learning methods to quickly and accurately predict dew point pressure. In this study, dew point pressure, the reservoir temperature, molar fraction (N2, CO2, H2S, C1, C2, C3, C4, C5, C6, C7+), relative molecular weight of C7+ (MWC7+), and relative density of C7+ (SGC7+) of 670 groups of condensate gas reservoirs in previous studies were collected, then Pearson correlation analysis was used to screen the main controlling factors, the selected features were put into the training of the model, the training set was trained by fivefold cross-validation, and finally different kernel functions were used to optimize the support vector machine model, and the conclusions were as follows: (1) The relative molecular weight of C7+ (MWC7+), the relative density of C7+ (SGC7+), the reservoir temperature (T (°F)) and the molar fraction (C5, C6, C4, C3, C7+) were highly correlated with the dew point pressure (2) Compared with the traditional linear regression, SVM (support vector machine) using Gaussian kernel has better prediction accuracy, and the obtained model can provide a reference for the prediction of dew point pressure in new condensate gas reservoirs.