Designing an automobile insurance classification system

Abstract

In automobile insurance, policyholders are generally classified according to rating factors such as age, territory and type of vehicle and charged appropriate premiums for their risk class. The problem of designing an efficient classification system can be considered as a contiguous clustering problem. A least squares decision criterion is employed and illustrated for both unidimensional and multidimensional cases. Both the benefit and cost associated with classifying risks are functions of the degree of complexity of the system, and the least squares criterion is used as a decision support mechanism for aiding system design decisions. © 1986.