Nature-inspired Preference Learning Algorithms Using the Choquet Integral
Michał Wójcik, Miłosz Kadziński · Proceedings of the Genetic and Evolutionary Computation Conference · 2024
We introduce various algorithms for learning the parameters of a threshold-based sorting procedure powered by the Choquet integral. This model accounts for interactions between monotonic criteria and facilitates categorizing decision alternatives into predefined, preferentially ordered classes. We focus on developing heuristic preference learning methods capable of efficiently processing large datasets of classification examples. Specifically, we utilize Local Search, Simulated Annealing, and nature-inspired approaches such as Genetic Algorithm, Fish School Search, and Particle Swarm Optimization. We demonstrate the effectiveness of the proposed model through a case study. Additionally, we present an experimental comparison of the recommendation accuracy achieved by these algorithms on a suite of benchmark sorting problems.