Finding better alternatives than those considered in a multiple criteria data sample

Abstract

This paper suggests a preference-based method to calculate the probability of finding a better alternative than the ones present in the original data sample. Such a procedure would be useful, for example, while selecting among job applicants, where each individual is evaluated in terms of multiple criteria. Once the evaluation of individuals is done, the efficient (non-dominated) individuals from the sample are identified and the most preferred candidate is chosen. However, it is often useful to assess the probability of finding even better candidates, were more applications invited. In this paper, we suggest a generic method to ascertain this probability, which can be used as a criterion to terminate the search for the most preferred alternative. This is achieved by estimating the value function of the decision maker in terms of different criteria. If a linear value function cannot be fitted, we suggest a novel approach to test more general functions in a stepwise manner.