The longshore current is a major and pervasive flow in the nearshore, and estimates of its velocity are needed in most coastal engineering projects involving sediment transport such as navigation channel maintenance and shore protection. In this study, support vector machine (SVM) is proposed to forecast daily longshore current velocities and at the same time the variability in performance of SVM with respect to the free parameters is investigated. The experimental comparisons between the SVM model and the classical back-propagation neural network (BPNN) demonstrate that the SVM is superior to BPNN in predicting longshore current velocities and has better generalization performance.