论文部分内容阅读
Generally speaking,pairwise comparisons are the most intuitive of the input data formats that can be utilized for group ranking problems.In most previous studies it has been the practice that user preferences are aggregated to form a full listing representative of consensus results.The fact that user options may be very discordant is ignored when preparing the total aggregated total ranking which makes decisions derived there from less than ideal.We have recently proposed a new concept,maximum consensus sequences,which finds of the longest ranking lists of items that agree with the majority and disagree with the minority.However,to determine the input requirement is still an arduous task,since the individual must provide a full ranking list for all items.In this study,not only do we relax this assumption by using a more flexible input format,specifically pairwise comparisons,but we also provide a new framework which can allow the inclusion of both total ranking and partial ranking data.An algorithm is developed to discover the maximum consensus sequences among all users while identifying conflicting items for which there is no user consensus.Finally,we illustrate the algorithm with a small numerical example.