Research On Personalized Recommendation Algorithm By Fusing Social Auxiliary Information

With the widespread use of the Internet and the emergence of a large number of intelligent terminals,the information resources that flooded in the network have exploded,and the information that is valuable to users,it is often overwhelmed by complicated and redundant mass information,and it is difficult for users to locate it effectively,which creates the problem of “information overload”.The recommendation system that designed to solve the problem of information overload is an effective tool.It aims to recommend things that may be of interest to users proactively.Collaborative filtering recommendation is the most widely used and successful among them.However,due to the problems of data sparsity and cold start,the quality of the recommendations generated by the system has become increasingly difficult to meet the actual needs of users.In recent years,the widely popular social network contains rich values information that can be used in the recommendation process,laying a material foundation for the system to form highquality recommendations.Therefore,this thesis uses collaborative filtering as a model to integrate information data in social networks effectively,and it designs corresponding algorithms in order to generate high-quality recommendations for users.The main work and innovations of this thesis are as follows:1.In view of the existing social information in the collaborative filtering recommendation algorithm fusion does not differentiate between the user's trust,and according to the related theory in sociology,the greater the influence of users in a social network to some extent,reveals the relevant opinions have been adopted more likely,therefore this thesis in order to establish a more accurate user social attributes,the establishment of the trust,communication and influencing factors into the research scope,and into the process to the recommendation of probability matrix decomposition model,is used to solve the recommendation system on sparse data sets to recommend the condition of the poor.Experimental results show that this method has a high recommendation quality on sparse data sets.2.Aiming at the problem that the existing trust model based only on scoring information and social relationship data is easy to cause distortion,the user's social tag information on the project is included in the research scope,and multi-source information is used to comprehensively analyze the user's explicit trust relationship and implicit Trust relationship,to establish a more accurate user trust model and project trust model,and apply it to the probability matrix decomposition model.Experimental results show that this method has good effect on the accuracy and diversity of recommendation.