The Research And Implementation On Group Recommendation Algorithm Based On User Preference Aggregation

The recommender system has developed extremely rapidly in recent years.The group recommender system,came into being based on multiplayer recommendation scenario has been greatly developed in recent years.In travel,movie and restaurant recommendation scenes,users are always spontaneous together to carry out activities,which need to be recommended for a group of people.Different from other single recommender systems,the group recommender system can not only recommend to a single user,but also can recommend a group composed of multiple users.The main contents of this paper are as follows:(1)Based on the traditional user preference aggregation method,an enhanced user preference aggregation method is proposed.It combines two basic user preference fusion methods and contains the aggregation features of recommendation aggregation and model aggregation.The experiment proves that "there is a correlation between personal preference and group preference",and applies the verification results in the improved method,which improves the recommendation accuracy.(2)Propose a group preference extraction algorithm and group representation algorithm based on classification information.This algorithm aims at the problem of fewer users rating items in a group.By generalizing the existing projects,the algorithm allows users to score the generalization results and achieve dimension reduction.Thus the algorithm alleviating the cold start and group scoring sparsity in the group recommendation system.(3)A mixed group recommendation prediction score algorithm based on classification information is proposed.The algorithm predicts the classification of the project.Its classification on the basis of the project and classification information based on the preference of the classification information group.The algorithm consists of two parts,one is the content-based recommendation algorithm based on classification information,and the other is the matrix decomposition collaborative filtering algorithm based on classification information.(4)Based on the innovative algorithm,combined with the actual needs of users,a group recommendation prototype system based on user preference aggregation is designed and implemented,which can display the recommended results and project contents effectively.By dynamically switching the preference aggregation method,the system chooses a aggregation method with lower time complexity to reduce the time overhead of the system when the recommended effect is not much difference.