Intelligent modeling of directed energy deposition contributes in improving the manufacturing accuracy of deposition. We designed an experiment with the deposition process parameters (laser power, powder feeding rate, scanning speed, and distance of the nozzle) as inputs, and the width and height of the cladding tracks as ouputs. A support vector regression (SVR) model based on radial basis function (RBF) was established to predict the size of the cladding tracks. Then, an improved particle swarm optimization algorithm was used to determine suitable values for the hyperparameters of SVR. Results indicated that the average relative errors of RBF-SVR for predicting the width and height of the cladding tracks were 4.58% and 5.33%, respectively, which were better than the results obtained using the back propagation (BP) neural network where the average relative errors were 6.72% and 7.96%, respectively. The RBF-SVR is suitable for predicting the size of cladding tracks and provides a reference for selecting process parameters in a directed energy deposition.