Group User Modeling In Group Recommendation Systems Based On Social Influence

With the widespread application of the Internet and the rapid development ofinformation technology, the amount of internet information has been experiencingexplosive growth, resulting in more time and more energy is needed to getting theuseful information. Faced with massive network resources, how to get access to usefuland accurate information efficiently has become the focus of research. Recommendersystems are techniques and software tools that suggest a carefully selected set of itemsto people, aiming to satisfy people’s preferences. As an effective way to overcome theinformation overload problem and increase people’s satisfaction, recommendersystems have been widely used in web-based applications.While much of the research on typical recommender systems focus on makingrecommendations to individuals, many daily activities involve groups of users.Besides, with the development of social network and online communities, more andmore activities have been carried out in groups online, the actual demand forproviding recommendation to group activities attracts more and more attention on theresearch of group recommender system.Group recommender system is expected to make suggestions that reflect thepreferences of the group as a whole, while resolving the disagreement among groupmembers and offering acceptable consensus options to everyone in the group.Recently, an increasing number of researchers have begun to study the new issuesrelated to group recommendation, and correspondingly, some remarkable grouprecommender systems have been developed.However, the shift of focus from recommending to individuals to recommendingto groups makes more of a difference than we might expect at first. Relative torecommendation to individuals, a number of new issues arise with the advent of grouprecommendations. Among all of the differences between individual recommendersystems and group recommender systems, one of the most significant differences isthe social factors between group members. Social factors like personality, expertisefactor, interpersonal relationships and preference similarities amplify the differencesbetween group recommendations and individual recommendations. Unfortunately,these factors have been ignored by most existing group recommender systems,leading to unsatisfactory recommendations for groups.In this paper, we revise state of the art approaches to group recommendation,focusing on the effect of social influence on group recommender systems. The novelty of this paper lies in describing a computational model designed for social influence ingroup recommender systems, integrating the influence of personality, expertise factor,interpersonal relationships and preference similarities. Besides, we have studied thedifferences of this method and traditional group methods.The experimental results show that the proposed methodology can provide highaccuracy and satisfactory group recommendations.