The Research And Implementation Of Group Recommendation Based On Graph Theory

With the development of society,social activities have become an indispensable part of people’s lives.However,people in the big cities live far apart,and their preferences will also be different.Busy work make each individual schedules vary a lot.All of these make people encounter a variety of problems when they make the choice of meeting place.The ideal venues should not only be able to meet everyone’s preference,but also meet their own time schdules.Therefore,how to satisfy the characteristics of the group to choose a meeting place while considering the above factors has become a focus of research content.In this paper,we analyze these several factors,including user’s preferences,time,distance and the integral relationship between the team respectively.On this basis,we propose a group recommendation algorithm based on graph theory.We improve the current rating prediction algorithm by preseting a full-rating model.This model use the hierarchical co-clustering algorithm to make user and item clusters,and then predict missed rating by combining groups Similarities and personalized preferences.Given the time and distance constraints,we design an item spliting algorithm.The splited items will be used to construct a bipartite graph and a one-dimensional projection graph.Finally,we obtain the final recommendation by using iterative HITS recommendation algorithm based on the paths calculated by the minimum spanning tree on the projection graph.By analyzing several experiments,each of the proposed algorithm performed better than the existing algorithms to a certain extent.Baed on the proposed algorithms,we design and implement a complete dinner recommendation system,divided into offline training and online recommendation.It enable users to complete the construction of group,make recommendations,exhibition and other related functions.Users can also obtain accurate recommendations immediately with the help of offline training module.A number of experiments confirmed that the algorithm can be userd in the real system...