POO-LPSP: New bio-inspired algorithm boosts AHP decision-making for AI vendor selection

Pairwise comparison matrices are the backbone of the Analytic Hierarchy Process (AHP), a staple tool for multi-criteria decision-making. However, Saaty's classic eigenvector method has long been questioned for its theoretical robustness in reflecting true priority vectors. This paper introduces a refined approach: the Least Penalty-Squared Prioritization (LPSP) optimization models, which include revised formulations like LPPDS and LPPWS to minimize RMPSV and RMPSWV. However, these non-linear models are computationally heavy for practitioners.

To solve that, author Kevin Kam Fung Yuen proposes the POO-LPSP method, integrating a Parallel Osprey Optimization Algorithm (POOA), a bio-inspired metaheuristic that efficiently navigates the complex LPSP search space. In a 16-page paper with 1 figure and 4 tables, the method is validated on a real-world Generative AI vendor selection problem, showing significantly lower computational overhead while maintaining or improving prioritization reliability. POO-LPSP thus emerges as a strong alternative to the traditional eigenvector method for AHP applications, especially where speed and accuracy are critical.