Self-Readjustment of Decision Model Parameters Based on Feedback: A Novel Approach for Dynamic Weighting

Real-world decision-making problems are inherently complex, requiring models that can accommodate evolving parameters, uncertain environments, and the subjective preferences of decision-makers, such as the importance weights assigned to various criteria. However, the subjective parameters often face two challenges: (1) they may be inaccurately set initially, and (2) they may change over time in dynamic environments. These issues can lead to mismatches between the decision model’s solutions and the decision-maker’s preferences. This paper introduces a novel approach, Feedback-based Readjustment of Parameters (FRP), developed for decision models whose output comprise a finite set of ranked solution alternatives. FRP leverages decision-makers’ feedback on the decision model’s proposed ranking of solution alternatives to iteratively infer and readjust the decision model’s subjective parameters. By integrating inverse optimization and an online learning mechanism, FRP provides an intelligent dynamic weighting approach. Numerical experiments on a multi-attribute decision-making ranking problem with randomly generated instances validate FRP’s performance and accuracy. The results demonstrate FRP’s effectiveness in aligning initial parameters with decision-makers’ preferences over time. This approach enhances decision support systems by ensuring adaptability to decisions feedback and responsiveness to dynamic decision contexts.