Research On Personnel Search And Recommendation Algorithms In Social Networks

The development of social network platform facilitates people's communication,and people can get the corresponding information through virtual network without leaving home.A large number of users publish information through social platforms to communicate,forming a complex social network.As the number of users in social networks increases,it becomes more difficult to accurately locate potential users in social networks based on the information input by users,and the accuracy of personnel search and recommendation also declines.In order to improve the accuracy of personnel search and recommendation,this paper combined user content,network topology and other factors to study the algorithm of personnel search and personnel recommendation.Finally,a personalized personnel search and recommendation system was implemented to provide users with search and recommendation services.The paper is divided into three parts,the specific research contents are as follows:(1)Research on personnel search algorithm based on Doc2 Vec and convolutional neural network.In view of the problem that users cannot be accurately found in social networks,this study built a user feature information model based on Doc2 Vec model and convolutional neural network,in order to deeply mine user information and carry out vectorization representation.The user feature information model was combined with Solr's original sorting algorithm,then the search results were sorted twice to meet the needs of target users for potential user discovery.Experiments show that the personnel search algorithm based on Doc2 Vec model and convolutional neural network improves 15.3% and 8.3% on Mean Average Precision(MAP)and 7.9% and 4.3% on Recall,compared with the search algorithm based on the single Doc2 Vec model and Solr search algorithm.(2)Research on personnel recommendation algorithm based on weighted LeaderRank.In view of the increasing number of social network users and the low accuracy of network recommendation,this study proposed a personnel recommendation algorithm based on the weighted LeaderRank algorithm by introducing chain-in and chain-out correlation,content correlation and time decay.This algorithm combined GloVe model,cosine similarity calculation method and Newton's cooling law to improve the deficiency of Weighted LeaderRank algorithm.The experimental results show that the precision of the proposed algorithm is higher than that of User-CF algorithm and User-Rank algorithm.Compared with the weighted LeaderRank algorithm,the algorithm improves 7.75% in precision,6.42% in click rate,and 4.27% in NDCG,effectively improving the recommendation quality.(3)Implementation of personnel search recommendation system.Combining the above proposed search algorithm and recommendation algorithm,combined with the current popular SMM(SpringBoot + MyBatis + MySQL)framework,CSS,JS and other technologies,the personalized personnel search recommendation system was designed to provide users with a concise and clear visual interface to better serve users and facilitate users to broaden their personnel.