Study Of Rough Support Vector Machine Algorithm Based On Fuzzy Method

Support vector machine (SVM) is a new machine learning technology and has drawnmuch attention on this topic in recent years. The theory of SVM is based on the VC dimensiontheory and structural risk minimization (SRM) principle. Currently, scholars have proposedmany improved SVM algorithms in the literature, for example, the fuzzy support vectormachine and so on. Nowadays, researchers begin to apply rough set methods to the researchof support vector machine and forces on the close integration between the rough set methodand support vector machine. At the same time, study of fuzzy method, rough method andsupport vector machine have become a new development direction.In this paper, support vector machine based on fuzzy method and rough method, namelystudy of rough margin support vector machine algorithm based on fuzzy method, has beenstudied. The main contents are as follows:First, on the basis of analyzing traditional support vector machine, fuzzy support vectormachine and rough margin support vector machine and aiming to existing noise sensitiveissues of the algorithms, the fuzzy weighted rough margin support vector machine model isobtained by introducing the fuzzy membership function and rough margin into the supportvector machines. By solving this optimization problem, fuzzy weighting based rough marginsupport vector machine algorithm is presented. It takes into account both the roughness ofmargin and fuzziness of the samples at the same time, that is to say, it not only considers thesample location in the hyperplane, but also considers the importance of the sample, so it canbe effectively reduce the impact of noise or outliers.Second, a fuzzy weighting based rough margin support vector machine for solvingmulti-class classification problem is proposed after analyzing support vector machinealgorithm and strategies based on one-to-one strategy and one-to-all strategy. The newalgorithm use fuzzy membership and rough margin, it effectively reduces the unclassifiedregions of traditional one-to-one strategy SVM and one-to-all strategy SVM.Third, aiming at two classification problem and multi-class classification problem; weselect several standard data set of UCI and so on. We study the performance of fuzzyweighting rough margin support vector machine by choosing the correct classification rate oneach sample set, the variance and the average correct rate on all the data sets. We compare theproposed algorithm’s performance with rough margin support vector machine, fuzzy support vector machine, and traditional support vector machine, respectively. To further reflect theperformance of presented algorithm, we use paired t-test and Friedman test to study theeffectiveness of the algorithm.