Finding Consistent Pairwise Comparisons with Genetic Algorithms
摘要
The paper provides a novel genetic algorithm for reduction of inconsistency in pairwise comparisons and then uses it for analysing inconsistency reduction for three popular scales, ten points scale, seven points scale and five points scale. The five points scale turned out to be the best, so we used it to analyze the effects of using different thresholds for differentiating between consistency and inconsistency. Furthermore, we compare the standard linear five point scale with two non-linear five point scales, that were introduced to connect qualitative pairwise comparisons with their quantitative counterparts.