A Bayesian optimization-support vector regression (BO-SVR) algorithm is proposed for the problem of predicting the molding density of press-loaded drugs. The orthogonal experimental method is used to collect the press-loading process parameters and quality data, and the data samples are analyzed for correlation, on the basis of which the support vector regression model is constructed, and the Bayesian algorithm is used to search for the optimal combinations of the penalty coefficients of the support vector regression model as well as hyperparameters such as the kernel function parameter, to evaluate the effects of the different combinations of the parameters, and to compare and analyze the effect of the BO-SVR model with that of the traditional support vector regression (SVR) model. The results show that BO-SVR nearly doubles all the evaluation indexes than the traditional SVR model.