Weighted Ordinal Support Vector Clustering

A weighted clustering method using the support vector machine approach is proposed for ordinal outputs problem. Based on the ideas of optimal hyper plane and nonlinear mapping, a linear clustering model in feature space is constructed which makes the margins between two separated groups maximal by solving a quadratic programming problem. And the affection of each training example to margins could be controlled by giving various weights of input data. As an application, the problem about regional food security division is solved by our algorithm. The result of experiment shows that it can deal with the unsupervised ranking learning problem effectively