Abstract: The forecast model based on support vector machine (SVM), which considers the water-absorbing capacity, dry density, wave impedance, Poisson ratio and elastic ratio as input index, and takes the compression strength as output index, was established. In order to increase the forecast accuracy of SVM, the particle swarm algorithm was used to optimize the SVM parameters, which can overcome the shortage for arbitrary selection of SVM parameters. Through analyzing the ore and rock sample data from Anqian mine and Gongchangling mine, the model obviously predicted more precise results than the nerve network method, with the biggest forecast error 8.2%. This result demonstrates that the method based on supersonic forecast strength is feasible and will be a new method for forecting rock strength.