Resumen de Weighted distance-based regression applied to automobile insurance data when the explosure base is not the unit

Eva Boj del Val , M. Mercè Claramunt Bielsa , Josep Fortiana Gregori

• We make use of the weighted metric scaling version of distance-based regression. We use this model for the selection of tariff variables in the rate making process. With this modification of classical multidimensional scaling we approach the case of grouped data. In the actuarial sense: When we have not individual claim experience; when all risk factors are categorical and we simplify model and computational time from the cross-classification; or simply, when we weight the studied units. Methodology is illustrated using a portfolio with real data from an Spanish automobile insurer.