A Group Decision-Making Method Based on Reciprocal Preference Relations Created from Sentiment Analysis.

Group decision-making is an event in which a group of experts have to decide between a limited set of alternatives. To share their opinions and ideas, this group of experts conducts a debate where they compare the alternatives with each other. Nevertheless, once the debate is over, the experts must provide their assessments by using reciprocal preference relations. This can be a problem, as an expert may not be consistent between what he expresses and his assessments. To create such consistency, in this work a group decision-making method is developed, which automatically creates the experts’ reciprocal preference relations from the comments they make in the debate. These comments are classified into positive and negative by using sentiment analysis techniques, specifically employing the majority membership in the bag of words of a given class. To calculate each value of the reciprocal preference relation, a new operator has been developed that uses the ranked comments and weights them. Nonetheless, the method offers the possibility to modify a reciprocal preference relation if the expert wishes to do so. Finally, this new group decision-making method develops a second operator that also uses the number of comments to adjust the weight of each expert. This operator quantifies the number of comments that an expert makes during the debate and gives him or her more weight based on the number of comments he or she makes. Consequently, the more comments the expert makes, the higher the weight he or she gets.