A clustering algorithm to find groups with homogeneous pre..

When we are learning people’s preferences the training material can be expressed as in regression problems: the description of each object is then followed by a number that assesses the degree of satisfaction. Alternatively, training examples can be represented by preference judgments: pairs of vectors (v, u) where someone expresses that he or she prefers v to u. Usually, obtaining preference information may be easier and more natural than obtaining the labels needed for a classification or regression approach. Moreover, this type of information is more accurate, since people tend to rate their preferences in a relative way, comparing objects with the other partners in the same batch. Starting form a set of preference judgments we have to learn a real ranking function f in such a way that f(v) > f(u) whenever v is preferable to u. If we accept linear ranking functions, f is represented by a hyperplane able to separate increasing or positive differences (like v – u when v > u) from decreasing or negative differences (like u – v); see [2, 3, 6]. We will employ an SVM classifier: SVM [4].