Research On Social Recommendation Algorithm Based On Rough Clustering

With the social economic developing and technology improving,the world meet a new era of Internet.With the social economic developing and technology improving,t the world ushered in a new era of the Internet.The Internet can offer a variety of services.Massive Internet information to provide convenience for users at the same time,also led to a series of problems such as information overload.How to get the most valuable information from the complicated information is the most urgent problem in the Internet era.Recommender system is one of the effective means to solve the information overload problem.Personalized recommender system can recommend the relevant information and commodities according to the user's preferences,and can recommend relevant services according to the user's social relations.Collaborative filtering is the most widely used technique in recommender systems,which makes recommendation for users by analyzing user ratings or other behavioral patterns.Despite the collaborative filtering recommendation technology has researched for many years,but there are still many problems such as data sparseness,cold start,low accuracy,etc.The proposed recommendation algorithm based on rough clustering can solve the problems of low recommendation accuracy and poor real-time performance.Firstly,rough set theory,K-means clustering algorithm and user-based collaborative filtering algorithm are combined to form rough K-means cluster recommendation algorithm.By analyzing the user's evaluation of information on items,to provide users with recommended services.In the recommendation process,the problem of attribution of boundary data is solved according to rough set theory,and the clustering effect is improved.Through the clustering algorithm,the search range of similar users is narrowed,the related calculation is reduced and the recommendation efficiency is improved.Secondly,the rough clustering algorithm is optimized from the viewpoint of data density.The density function of each data is calculated,and the data with the largest density and the farthest distance are chosen as K initial clustering centers to overcome the shortcomings of randomly selecting the initial clustering center in the rough clustering algorithm.By calculating the degree of influence of each data on the clustering,the weights of the rough K-means clustering approximation set are self-adaptive,and the rough clustering results can avoid the influence of the preset weight.At the same time,based on the idea of Weighted Slope One,this thesis proposes a user item rating adjustment method to eliminate the effect of rating bias on the recommendation effect in user history rating data.Finally,the rating adjustment method and the rough clustering recommendation algorithm based on density improvement are combined to comprehensively improve the recommendation effect.The effectiveness of the proposed algorithm is verified by the comparison experiment.The experimental results show that the proposed algorithm can solve the problem of low accuracy and low real-time performance of traditional collaborative filtering algorithms.And the proposed algorithm makes a useful attempt for the concrete application of the socialized recommendation algorithm.