Mineração de preferências contextuais fuzzy

Abstract

In recent years, much researches in Preference Mining has focused on development of methods for mining a preference model from crisp pairwise representation. In this research, is presented the FuzzyPrefMiner algorithm, developed for mining fuzzy contextual preference model from fuzzy preference relation. This type of preference representation is composed by a set of triples (u; v; n), where, u and v are tuples evaluated by user and n is the degree of preference of tuple u with respect to v. The fuzzy preference relations enable the study of the consistency of users and in this context, are presented two methods for repair inconsistency of these relations: No Incremental Range Voting and Incremental Range Voting. Both methods are based on technical Voting System and are used for repair the inconsistency of fuzzy preference relations inferred by FuzzyPrefMiner. A set of experiments in real data were performed to validate the FuzzyPrefMiner algorithm and the methods for repair inconsistency. The results quite satisfactory when compared with the state of the art regarding Preference Mining Algorithms and Methods for Repair Inconsistency.