Influence, Relevance and discordance of Criteria in AHP-Global Bayesian prioritization

This work extends the AHP-Bayesian Prioritization procedure proposed by the authors for a local context (a single criterion) to a global context (a hierarchy) and presents a taxonomy of criteria that contribute to the ranking of the alternatives. In this global context, the paper defines and characterizes: (i) the influence of a criterion in the final prioritization of the alternatives by means of cross-validation techniques; (ii) its degree of discordance with the rest of the criteria of the hierarchy utilizing new discrepancy measures; (iii) its relevance in the final decision. Along with the taxonomy and its corresponding methodology, the paper introduces new decisional tools for the establishment of the classification and demonstrates their use from a cognitive perspective (knowledge extraction) whilst paying special attention to the resolution of the P.α and P.γ problems. The proposed methodology is illustrated by a case study on an e-participation process developed for the Zaragoza City Council (Spain) and implemented by the Zaragoza Multicriteria Decision Making Group.