The number of support vectors(SVs) of support vector machine(SVM) is usually large, which results in a
substantially slower classification speed than many other approaches. The less SVs means the more sparseness and the
higher classification speed. Therefore, an algorithm called FD-SVM is proposed, which employs FCM to cluster a dense
SVs set to a sparse set. Then, aiming to minimize the classification gap between SVM and FD-SVM, a linear least square
programming model is built for obtaining the optimal coefficients of the new sparse SVs. Experiments on toy and real-world
data sets demonstrate that, after reducing 50% of SVs, an increase of 50% on the classification speed is achieved, while the
performance of losing maintains a statistically insignificant level.