Abstract: The structure information in data has not been exploited in the Least Squares Support Vector Machine Classifier(LSSVM ) and the LSSVM is sensitive to the outliers．Focused on the above issues of the LSSVM, this paper proposed a new classifier---structural weighted least squares support vector machine (SWLSSVM)．The structure information is considered by incorporating the covariance matrix into the objective function,and in order to reduce sensitive to the outliers,according to difference of the distances from different types of samples to the center of the sample,the different weights are assigned to the different training samples in the error term of objective function．The experimental results show that the SWLSSVM is more superior to the LSSVM and the SVM in classification and generalization performances.

Key words: Least squares support vector machine,Structure,Weight,Covariance matrix