Hide/Show Apps

A probabilistic approach to multi criteria sorting problem

Buğdacı, Aslı Gül
We aim to classify alternatives evaluated in multiple criteria among preference ordered classes assuming an underlying additive utility function. We develop a probabilistic classification method by calculating the probability of an alternative being in each class. We assign alternatives to classes based on threshold probabilities. We require the decision maker to place an alternative to a class when no alternatives satisfy the required thresholds. We find new probabilities for unassigned alternatives in the light of new information and repeat the procedure until all alternatives are classified. We implemented our algorithm to classify MBA programs among preference ordered groups. We evaluate our algorithm based on the number of misclassified alternatives and the number of alternatives placed by the decision maker.