Data selection for support vector machine classifiers

The problem of extracting a minimal number of data points a large dataset, in order to generate a support vector (SVM) classi er, is formulated as a concave minimization and solved by a nite number of linear. This minimal set of data points, which is the number of support vectors that completely characterize separating plane classi er, is considerably smaller that required by a standard 1-norm support vector machine or without feature selection. The proposed approach incorporates a feature selection procedure that in a minimal number of input features used by the er. Tenfold cross validation gives as good or better results using the proposed minimal support vector machine (MSVM) classi er based on the smaller set of data compared to a standard 1-norm support vector machine er. The reduction in data points used by an classi er over those used by a 1-norm SVM classi er 66% on seven public datasets and was as high as 81%. This makes MSVM a useful incremental classi cation which maintains only a small fraction of a large dataset merging and processing it with new incoming data.