Machine Learning-Driven Approach for Large Scale Decision Making with the Analytic Hierarchy Process

The Analytic Hierarchy Process (AHP) multicriteria method can be cognitively demanding for large-scale decision problems due to the requirement maker make pairwise evaluations of all alternatives. To address this issue, paper presents an interactive that uses online learning provide scalability AHP. proposed involves a machine algorithm learns maker’s preferences through small subsets solutions, and guides search optimal solution. methodology was tested on four optimization with different surfaces validate results. We conducted one factor at time experimentation each hyperparameter implemented, such as number alternatives query maker, learner method, strategies solution selection recommendation. results demonstrate model is able learn utility function characterizes in approximately 15 iterations only few comparisons, resulting significant cognitive effort savings. initial subset solutions chosen randomly or from cluster. subsequent ones are recommended during iterative process, best strategy depending problem type. Recommendation based solely smallest Euclidean Cosine distances reveals better linear problems. also easily incorporate new parameters methods comparisons.