Jonas Daniels


Preference Modelling
Learning the user’s subjective choice

Abstract

A preference model is an algorithm that is supposed to help the user make choices. This project implements a preference model for a user’s choice of movies. Is it possible to apply a preference model to a user’s choice of movies? The implementation of the algorithm was made with the simplex algorithm as base. The simplex algorithm tries to maximize a given function subjective to some given constraints. The constraints that are given to the simplex algorithm are previous selections in movies made by the user. The algorithm was tested having the test subjects make 15 movie selections and then ordering a list of five movies by preference. The 15 choices became constraints and the preference model calculated its own list. The two lists were then compared in order to see how well the preference model did. In 30 of the 32 test cases the algorithm picked the same first choice as the user, but it did not do so well on the rest of the list. Objective values for movie are easiest to find if they are true/false values and the algorithm would most likely have done better if true/false values were used. The algorithm implementation can be used for any other area where items can be viewed as objective values on a scale.