Instances Reduction Support Vector Machine Base On Rough Set

Support vector machine (SVM) and rough set are hot research in the field of artificialintelligence and machine learning today. Support vector machine is a novel approach forpattern classification rooted in statistical learning theory, the principle of structural riskminimization is used as the criterion of classification, and the optimal classificationhyperplane is constructed from support vectors near the boundary. Only the support vectorshave contribution to classification. However, solving SVM is based on whole training set.When the training set is very large, it will require a great amount of memory and take a longtime to search the optimal solution.In order to deal with the problem mentioned above, this paper presents a method namedsupport vector machine based on instance reduction with tolerance rough set. The idea of thealgorithm is that almost all support vectors are nearby the boundary of classification. So wecan employ tolerance rough set method to find the boundary region (BR), the instances in BRcan be used as candidate support vectors to train SVM. Furthermore, rough set can alsoremove redundant attributes by attribute reduction keeping the ability of classification.Therefore, the method proposed in this paper can simultaneously reduce attributes andinstances. Specially, firstly, attributes and instances are reduced by use tolerance rough set,secondly, the SVM is trained from the reduction instances. In addition, an attribute reductionmethod was proposed in this paper based on instance selection. The experimental results showthat the proposed method is effective and can efficiently reduce the computational complexityboth of time and space especially on large databases.