Research On Recommendation Algorithm Based On The Social Influence Of Neighbors

With the development of the Internet,there are more and more Internet users,at the same time,the number of online information increasing at an exponential speed.So,it also led to information overload problem.When faced with the vast amounts of information,the users are often difficult to accurately and quickly to find the valuable information which they are interested,and the personalized recommendation system is one of the most effective tools to solve this problem.The traditional collaborative filtering recommendation method mainly depends on the users of the product evaluation information,in addition,with the popularity of various popular social networks and intelligent terminals,the user's social network information has also been used in personalized recommendation technology and it Used to make up for the inadequacy of sparse environment rating information.But at present,most of the research focuses on trust relationships in a social network,While it ignoring the different influence of friends on users of the final decision also plays a different degree of influence.In the social networking site,in addition to user to user interests and preferences because of relatively strong performance,in order to put them outside the cluster,social impact is also an important feature.Its impact on the social network user behavior may be more powerful than the interest of the user and user.This paper introduces the background and significance of the research on the social influence of the neighbors in recommendation algorithm,and then discuss the current research on recommendation system,the social influence of neighbor-based recommendation system and the based on the network structure recommendation system,it summarizes the current existing research results.Then introduces the basic knowledge of the recommendation system,from three aspects to research the influence of neighbors on social network,they are the users penetration,the user's degree and the shortest distance between user and user,proposed recommendation algorithm based on the social influence of neighbors.And then,combines the social influence of neighbors and the mass diffusion based on initial resource allocation,for the target users,find the top N greatest influence on him.When the mass diffusion of initial resource allocation,the top N greatest influence users' score to the item which the target user had selected as the weight of resource allocation,and when the user spread to item also used this idea,non-uniform distribution of user resources.Finally,in this paper,I use the Epinions dataset on experiment for the social influence model and the combination algorithm between the social influence model and mass diffusion.By comparison with the traditional collaborative filtering recommendation algorithm,recommendation algorithm based on item popularity and the mass diffusion algorithm based on uniform allocation of resources,analysis the recommendation accuracy rate,the weights of ranking,intra-user-diversity,inter-user-diversity,novelty and coverage on six areas,and then discovery the social influence of neighbor-based recommendation algorithm and the mass diffusion base of social influence of neighbor recommendation algorithm can let the performance of recommendation system has been improved.