Data-driven modeling is considered as a prospective approach for many conventional physical problems including ocean applications. Among various machine learning techniques, support vector machine stands out as one of the most widely used algorithms to establish models connecting pertinent features to physical quantities of interest. This paper takes the experimental data for a fixed cylinder in shallow water as the baseline data set and explores the modeling of nonlinear wave loads by the support vector machine (SVM) regression method. Different feature and target selections are studied in this paper to establish the nonlinear mapping relations from ambient wave elevations and kinematics to nonlinear wave loads. The performance of the SVM regression model is discussed and compared with nonlinear potential flow theory focusing on the overall statistics (standard deviation and kurtosis), which is critical for fatigue and extreme statistics analysis.