Research And Implementation Of Group Movie Recommendation Algorithm Based On User Preference Fusion

Group recommendation plays a more and more important role in real life.Group recommendation is not only a simple addition of individual recommendation results,but also needs to consider group characteristics.The core of group recommendation is to consider the interest preference information of all members in the group,and get the final recommendation result through recommendation algorithm and fusion strategy.For example,when friends go to dinner together,we should consider everyone’s opinions and finally choose the dining address.The most important goal of group recommendation system is to recommend satisfactory results for groups.Traditional recommendation models,such as logistic regression and matrix decomposition,are static and cannot dynamically adjust the user’s weight index in the process of model training.They all fuse preferences through predetermined preference fusion strategies.These methods and strategies ignore the dynamic changes of user preference information,do not consider the dynamic interaction information between users,and can not meet complex recommendation scenarios,Therefore,the recommendation effect of this static recommendation model is often not very good.In view of this situation,this paper proposes two improvements,one is based on the traditional recommendation fusion method,the other is the group preference fusion model based on attention mechanism.The main research contents of this paper are as follows:(1)The first point is to fuse user and group information,that is,recommendation fusion and model fusion.Model fusion is mainly for group recommendation,and the final recommendation list is generated through group recommendation model and algorithm.Recommendation fusion is mainly for individuals,generating each person’s recommendation list,and then using the fusion strategy to fuse the individual recommendation list.Finally,the recommendation list generated by model fusion and recommendation fusion is predicted through certain rules to generate the final recommendation list.(2)The second point is to use the attention mechanism and the group preference fusion model based on the attention mechanism to dynamically learn the relationship between users and groups.In the preference fusion stage,the group preference information and the user’s personal preference information are fully considered,so that the model always considers the user’s personal relevant information in the whole process of fusion.In the prediction stage,user item vector and group item vector share the same hidden layer structure,and multi task learning is introduced to make multiple tasks train in parallel,learn from each other and strengthen each other.This paper enables user item vector and group item vector to promote each other,and the weight indexes of groups and individuals can be continuously adjusted in the process of training.The experimental results show that the proposed method improves the accuracy of group recommendation.