Research On Multi-granular Hesitant Fuzzy Linguistic In Formation And Their Applications In Group Recommendation

A group recommended system (GRS) is an information system which suggests items for meeting the common interests of a group of people engaged in a group activity. Though there have been some studies on group recommended systems, they focus on off-line environments and established groups. These days, many group activities and interactions are done in a virtual space,so research on GRS for virtual communities in online social networks will become a hot issue in the recommended systems field. As the group recommendation methods are the key problems of the group recommendation system, it is of major theoretical significance and application value to study the group recommendation theory and group recommendation methods.In the group recommendation system, when the individual is used to describe the preference information,natural information is utilized. Because the preferences of different individuals are different, so the preference information usually has the characteristics of fuzziness and hesitation. In addition, due to the different network platforms may be used to describe the discrete granularity of language information, so the group preference information also has multi granularity. Take into account these,in this thesis,we study the method of user group discovery in online social networks.Secondly, on the basis of studying the individual score prediction, this thesis deeply studies the methods of group users' preferences in online social networks. Finally,the TOPSIS group recommendation method and VIKOR group recommendation method for online social network users are studied. Specific research work and innovations are as follows:(1) Group discovering method in multi-granular hesitant fuzzy linguistic environment. The group discovering is the basic problem in group recommendation system. First of all, on the basis of the multi-granular hesitant fuzzy linguistic term set,this paper introduces the formula of the cosine similarity of the multi-granular hesitant fuzzy linguistic term set. The cosine similarity formula is utilized to calculate the similarity between users. Secondly, the minimum spanning tree method is extended to the multi-granular hesitant fuzzy linguistic environment for clustering analysis.Finally, the minimum spanning tree clustering method and the equivalence relation clustering method is compared, and the rationality and validity of the proposed method are analyzed.(2) The method of scoring prediction under the environment of hesitant fuzzy linguistic set. Score prediction is one of the focuses of the research of a group recommendation system. In this part, the principal methods of scoring prediction in group recommendation system are summarized. Secondly, the distance similarity formula, cosine similarity formula and the similarity coefficient formula are proposed to calculate the similarities between users. Finally, the distance similarity formula, the cosine similarity formula and the correlation coefficient similarity formula are utilized to calculate the unknown information, and the accuracy of the three methods is compared. The feasibility and validity of the proposed correlation coefficient similarity formula of hesitant fuzzy linguistic information score are analyzed.(3) Group preference aggregation method for multi-granular hesitant fuzzy linguistic term sets. It is a fundamental problem how to aggregate individual preference information into group preference information in group recommendation system. For this reason, we first put forward the concept of triangular hesitant fuzzy sets and study the properties of triangular hesitant fuzzy sets. Multi-granular hesitant fuzzy linguistic term sets are transformed into triangular hesitant fuzzy sets. Secondly,we define the generalized triangular fuzzy weighted averaging operator and generalized triangular fuzzy weighted geometric operator, and derive the properties of the two operators. Finally, taking the automobile recommendation as an example, the two operators are utilized to aggregate the group preference described by the multi-granular hesitant fuzzy linguistic information.(4) TOPSIS group recommendation method for multi-granular hesitant fuzzy linguistic set. According to diverse groups of preference information with multi granularity, hesitant fuzzy characteristics, this thesis firstly defines the concept of multi-granular hesitant fuzzy linguistic set,series distance formulas of multi-granular hesitant fuzzy linguistic sets are defined. The properties of these formulas are studied and the relationship between them is considered. Secondly, under the condition that the attribute weights are completely undetermined, the goal programming model is established, and the attribute weights are obtained by solving the Lagrange equation.In the case of incomplete attribute weights, the linear programming model is utilized to solve the attribute weights. Finally,distance formulas and TOPSIS method are used to group recommendation, and the influence of the parameters of the formulas on the satisfaction and recommendation results is analyzed.(5) Group recommendation method of VIKOR with multi-granular hesitant fuzzy linguistic information. Frist, based on the concept of multi-granular hesitant fuzzy linguistic term sets (MHFLTS), the entropy measures are defined, and the attributes of recommended items can be calculated by using MHFLTSs' entropy measures.Secondly, the extended VIKOR method of MHFLTSs is proposed, and the extended VIKOR method is utilized in group recommendation. Finally, this thesis discussed the differences between the VIKOR method and the TOPSIS method through theoretical analysis, numerical experiments and sensitivity analysis. The group recommendation example verifies the effectiveness of the proposed method.In a word, based on the multi-granular hesitant fuzzy linguistic environment, the online social network group users discovering method, group users' preferences acquisition method and group recommendation method, the strategic problems of the group recommendation systems are studied deeply and systematically. The research results not only expand the theory of fuzzy mathematics, and has guided significance for the group recommendation system and other group decision making problems.