The process is based on iterations of the three steps, which are herein detailed.1. Pair Sampling. An automated procedure selects from the set of Requirements a set of Sampled Requirements Pairs whose relative preference is unknown (i.e., Unordered Requirement Pairs, as defined in (1)), according to a sampling policy. A sampling policy can be a random choice or it may take into account the rankings computed in the previous iteration.32. Priority Elicitation. This takes the collection of Sampled Requirements Pairs produced by the Pair Sampling step in input and produces as output a set of Ordered Requirements Pairs on the basis of the Priorities expressed by a decision maker.3. Priority Learning. Given a partial elicitation of the stakeholder priority and eventually a set of Ranking Functions, the learning algorithm produces an approximation of the unknown preferences and then the correspondent Approximated Rank for the requirements.The Approximated Rank, that is, the output of the process, represents an approximation of the exact ranking and may become the input for a further iteration of the process. If the result of the learning step is considered accurate enough (or time to input preferences runs out), the iterations stop and the process gives the last approximated rank (Final Approximated Rank) as output.
đang được dịch, vui lòng đợi..
