Detecting Community Based On Improved Particle Swarm Optimization And Friends Recommendation System Research

With the rapid development of Internet technology,the resource of network becomes richer,but the massive information can not bring better experience for the user.On the contraty,it becomes more difficult.The negative effects causing by excessive information is called information overload.Scanning a large number of irrelevant information must make the customers continual loss.So recommendation systems emerge as the times require.At present,as one of the classic applications,friends recommendation becomes an important and practical personalized service of social networks,it plays a key role in promoting the development of social network.However,the low recommendation accuracy and the long recommended time hinder the further development of the recommendation system.The purpose of this paper is to solve these problems.As the important characteristics of social network,the community structure has a profound influence on the effect of the recommendation system.The division of community structure is to classify users into different sets,so that internal node has a higher degree of similarity,and different sets have a heavy dissimilarity.In essence,community division is research on an attribute of picture collection,the result is a set of optimal value.Combining the thought of community division and the recommendation algorithm will bring new inspiration and breakthrough inevitably.As follows are the improvements of this paper: first,it brought the thought of community division into friend recommendation,so that users who will be recommended can be solved into different social circle.Then calculating the similarity of these from the same community,it can reduce similarity calculation and improve the accuracy of the recommendation;in addition,the paper combined Particle Swarm Optimization algorithm with community division and improved it,after that,Suboptimal Experience Particle Swarm Optimization algorithm is proposed.In the improved algorithm,each iteration should consider not only its optimal fitness values and the global best fitness value,but also the impact of the suboptimal particle,so that it can avoid the precocious phenomenon;finally,it introducted the similarity of user's personal information into the recommendation,and put the user comprehensive similarity as the recommended standard.The results that SEPSO algorithm compared with the standard Particle Swarm Optimization algorithm show that: SEPSO algorithm can effectively improve the classification accuracy,and community division structure is more obvious.It is compared with the user link similarity and user comprehensive similarity,the results show that: the proposed algorithm has higher recommendation accuracy,it is better than standard algorithms.