Weighted Support Vector Machine Based Clustering Vector

In order to set the weights of weighted support vector machine, the method that based the clustering vector was proposed in this paper. The clustering vector was brought in and the weights were set by calculating the distances between the data of every class and their clustering vector. In SVM, the data were mapped into a higher dimensional feature space and the distribution of data was changed, so the clustering vectors and distances were calculated in the feature space. It was confirmed by simulation that the rate of accuracy was promoted in this way, and the classification accuracy for the class with small training size was improved for the uneven training data. The effect of outliers can also be alleviates by using this method.