Fuzzy Multiclass Weighted Support Vector Machine

in this paper, we propose weighted support vector machine (WSVM). Dividing input space to several subspaces gives us properties of high dimensional space. In each subspace a SVM is trained and resultsfused based on portion of input sample to each subspace. Dominant points of WSVM are, increasing dimensionof input space, focusing of SVM ability in each subspace, and combination of SVM classifiers in decision making. For multiclass case, we introduce fuzzy version of multiclass SVM over WSVM namely FWSVM. Obtained results over synthetic samples and iris data show superiority of the proposed WSVM and FWSVM relative to SVM and FSVM.

Keywords:

Support vector machine , high dimensional space , classifier combination , fuzzy multiclass