Notice that the first and the third steps of the process are automatic. Concerning the second step, we will assume—for simplicity—that the preference elicitation is monotonic (i.e., the decision maker does not evaluate the same pair twice). Using the concepts introduced in the previous section, the entire prioritization process can be conceived as anapproximation problem where, given a finite set of requirements Req and a set of pairwise priority relations between requirements of the kind ri rj specified by the stakeholder, the challenge is to learn the Final Approximated Rank HðrÞ such that 8ri;r j we have HðriÞ >H ðrjÞ if KðriÞ >K ðrjÞ, where KðrÞ is the unknown target prioritization ranking (Target Rank). The objective is twofold: to minimize the elicitation effort, while reducing the disagreement between the target (K) and the approximate rank (H).
đang được dịch, vui lòng đợi..