Improved SVM classification algorithm based on KFCM and LDA

Abstract

To address the problem that SVM is sensitive to outliers and noise points, in order to improve the classification accuracy of SVM, this paper introduces fuzzy theory and intraclass dispersion theory, proposes an improved SVM classification algorithm, uses KFCM and LDA to filter the data set, and selects reasonable training samples, thereby reducing the number of wild points and noise points in the training sample, and thus reducing its impact on the classification effect of the classification model. Compared with the traditional SVM, the algorithm in this paper considers the impact of training samples on the classification effect, introduces fuzzy theory and intra-class dispersion, and eliminates the wild points and noise points in the training samples that affect the classification accuracy of the classification model. Experimental verification shows that the classification accuracy of the SVM classification model trained by the filtered training samples is higher than that of the SVM classification model without the trained training samples.