Starting from the assessment itself, the construction method of learning samples for less than the traditional method of sensitivity analysis is proposed based on the operational effectiveness of SVM sensitivity analysis. Analysis based on least squares support vector machine (Least Square Support Vector Machine, LS-SVM) operational effectiveness model and principle. The paper proposed a new learning sample construction method based on attribute utility function estimation, and PSO algorithm LS-SVM parameter optimization method. Summary based on the operational effectiveness of support vector machines algorithm sensitivity analysis, using the data in reference example for numerical example. The results show that the method does not need to consider the complex relationship between the decision attribute, in the flexibility and the amount of information provided on the conventional analytical method, calculation speed is better than the Monte Carlo method.