A fast fuzzy support vector machine based on information granulation

摘要

In order to improve the efficiency of fuzzy support vector machine training high-dimensional and large-scale dataset, a fast fuzzy support vector machine based on information granulation (FSVM-FIG) is proposed. Firstly, the training set is divided into some granules by fuzzy C-means, including pure granules and mixed granules. Since most support vectors are close to the border of two classes of samples, we believe that the support vectors must be in mixed granules, so we save only the mixed granules for new training set. In addition, because there are some noises and outliers on the border of two classes of samples, we use the k-nearest neighbor algorithm to remove noises and outliers. Finally, we use fuzzy support vector machine based on cluster hyperplane to train the final training set. Experimental results show that FSVM-FIG can not only improve the training efficiency of the training sets that contain noises and outliers, but also ensure a certain degree of prediction accuracy.