Learning of a Decision-Maker's Preference Zone With an Evolutionary Approach.

Area of Science:

• Artificial Intelligence
• Optimization
• Decision Making

Background:

• Multiobjective optimization problems involve trade-offs between competing goals.
• Decision makers often struggle to articulate precise preferences in complex spaces.
• Existing methods may require extensive preference elicitation.

Purpose of the Study:

• To develop an evolutionary learning algorithm for identifying a decision maker's preferred solutions in multiobjective spaces.
• To enable efficient learning of an ideal point based on pairwise comparisons.
• To provide a robust method for decision support in conflicting optimization scenarios.

Main Methods:

• An evolutionary learning algorithm utilizing pairwise solution comparisons from a decision maker.
• Learning an ideal point that guides the evolutionary process towards preferred solutions.
• Iterative refinement of solutions based on decision maker feedback.

Main Results:

• The proposed algorithm accurately identifies the decision maker's preferred solution zone.
• Effective learning is achieved with a small number of preferences and in few generations.
• The method demonstrates robustness against inconsistent preference statements.

Conclusions:

• The evolutionary learning algorithm offers an efficient and effective approach for multiobjective decision making.
• Pairwise comparisons provide a practical mechanism for preference elicitation.
• The algorithm's robustness makes it suitable for real-world applications with potential data inconsistencies.