Abstract: In the traditional calculation of sulfur capacity, some parameters in the mechanism model are difficult to obtain. To solve this problem, a regression method was proposed based on the AdaBoost and LS-SVM approaches. Sulfur capacity can be predicted by this method. In the method, LS-SVM has fast computation and it is suitable for the problems with small sample set. The AdaBoost method can combine the weak learning machines into a strong learning machine and its accuracy is higher than single LS-SVM's. Meanwhile, the method can reduce the effect of parameters on final prediction results. The result of the simulation shows that this method can significantly improve the prediction accuracy and meet the production requirement.