Weighted Fuzzy Support Vector Machine And Its Applications

Support Vector Machine(SVM)is a machine learning method based on statistical learning and optimization theory.Because it resolves the local minimum,small sample and nonlinearity problems perfectly,it has become a hot field and drums up the development of machine learning.Traditional SVM is very sensitive to noises and outliers in the training samples which can affect the classification performance.Withal,scholars at home and abroad construct Fuzzy Support Vector Machine(FSVM)by introducing the fuzzy membership function to SVM.It takes into account of the influence of different input samples on the acquirement of the optimal hyperplane and improves the robustness of the classification effectively.Fuzzy membership is the key point as well as nodus because of the direct affects to the performance of FSVM.And there is not an uniform standard.In this thesis,we studied the algorithm of FSVM,and made many improvements.Then the improved FSVM are verified by experimental data.Our main work has been done as follows:1.Do many detailed studies about SVM and FSVM,including the generation of models,derivation,evolution and their nonlinear promotion.On the part of FSVM,the definition and theorem of fuzzy mathematics in fuzzy support vector machine are given.2.A method for constructing membership function based on the distance between sample and hyperplane is proposed.Then the membership is introduced into the FSVM model and a weighted fuzzy support vector machine based on the distance from hyperplane(IFM-WFSVM)is proposed.The effectiveness of the proposed algorithm is verified on the artificial datasets and UCI datasets respectively.3.The other method for constructing membership function based on DP Clustering is proposed.Then the membership is introduced into the FSVM model and a weighted fuzzy support vector machine based on DP clustering(DP-WFSVM)is constructed.The experimental results and performance analysis are given to show the good classification performance.In this thesis,the performance of FSVM is improved by using the theory of statistical learning,fuzzy mathematics,FSVM,DP clustering and the properties of hyperplane.These studies are of great significance in improving the classification performance.