CPP-TRI: a sorting method based on the probabilistic composition of preferences

This paper develops an approach for sorting based on randomising the trichotomic decision about classifying an alternative as superior, inferior or indifferent to a certain number of profiles. The alternatives to be classified and the representative profiles are determined by performances under multiple criteria. The alternative is allocated to the category for which the probabilities of being above and below such a category are nearest to one another. A previous development in this stream is ELECTRE TRI-nC. Results of application of the new approach are compared to those obtained in a previous application of ELECTRE TRI-nC. A second analysis explores the new method when applied to a situation in which the evaluations come from a group of people instead of only one evaluator. The robustness of the approach was also tested by comparing classifications from benevolent and exacting rules. The results of application of the different rules are similar.