A comparison of support vector machines and traditional techniques for statistical regression and classification

[摘要] ENGLISH ABSTRACT:Since its introduction in Boser et al. (1992), the support vector machine has become apopular tool in a variety of machine learning applications. More recently, the supportvector machine has also been receiving increasing attention in the statisticalcommunity as a tool for classification and regression. In this thesis support vectormachines are compared to more traditional techniques for statistical classification andregression. The techniques are applied to data from a life assurance environment for abinary classification problem and a regression problem. In the classification case theproblem is the prediction of policy lapses using a variety of input variables, while inthe regression case the goal is to estimate the income of clients from these variables.The performance of the support vector machine is compared to that of discriminantanalysis and classification trees in the case of classification, and to that of multiplelinear regression and regression trees in regression, and it is found that support vectormachines generally perform well compared to the traditional techniques.

[发布机构] Stellenbosch University