Research On User Relationship Strength Estimation Model In Online Social Networks

With the development of the Internet,online social networks are becoming popular rapidly and play an important role in people's daily life.Which open up an important channel for the action of information dissemination,experience sharing,and life experience exchange.Therefore,relationship strength between users in online social networks has attracted great attention from researchers.Social networks have gradually become an important role in many fields.Which can be applied to fields,such as friend recommendation,product recommendation and link prediction?For personalized service recommendation,the relationship strength between users in online social networks is an important basis for making recommendations.The preferences of target user who has recommended requirements tend to be more close to the user who has the strong relationship strength with him,and the recommendation from a more intimate person tend to be more easily accepted.Thus,it is obvious that the relationship strength between users in online social networks is important.But the existing relationship strength calculation researches are unilateral.Many methods are just a general way to calculate the relationship between the users.It only considers direct relation and ignores the importance of indirect relation.Which lead to inaccuracy of relationship strength estimation as a result.In view of the shortage of the present study,this paper shows a user relationship strength estimation model in online social networks based on fusion of activity field classification and indirect relationship.And the main research contents include the following aspects:Firstly,crawl the relevant data from the social network,then make data preprocessing(including word segmentation processing and remove stop word),and transform raw data into the corresponding document data set.The removal of garbage data is useful to improve the accuracy of the calculation results.Secondly,make activity field classification for interactive activities of user groups in the online social network.Use LDA algorithm to cluster user interactive activities document,and utilize the normalized Google distance to calculate the similarity between the result cluster and activity field names(including job,diet,shopping,travel,sports and entertainment),and select the activity field for each result cluster.After that,select the activity field for each interactive activity document through the similarity calculation.Combining the activity field classification to calculate the relationship strength between users in the online social network can make the research result be more targetedly applied to other fields in the future.For example,as for personalized recommendation,we can make recommendations based on different field to improve the probability of success.Thirdly,various factors are considered in the calculation of direct relationship strength.The individual similarity,timeliness,interactivity are used to calculate the direct relationship strength between users in each interactive activity field.Which cover the influence factors in various aspects of relationship strength between users in online social networks,and it is conducive to the accurate calculation of the direct relationship strength.Fourthly,the indirect relationship is considered in the relationship strength calculation.The author fuses the indirect relationship with the calculation of the final relationship strength due to the importance of indirect relationship in online social network,which solves the problem that there is no direct relation between the users lead to the relationship strength cannot be calculated.So,it can improve the accuracy of relationship strength calculation.Fifthly,the author comes up with an evaluation indicator to measure the accuracy of the relationship strength calculation results.The efficiency of activity field classification method is evaluated by comparing this method with other activity field classification method including document level,cluster level,micro-blog session.In addition,the relationship strength estimation result is measured based on Precision,Recall and the normalized Discounted Cumulative Gain(nDCG),and comparing the method of this paper with linear combination method and general framework model method,the experimental results show that the method of this paper is more accurate.