User Influence Evaluation Of Social Network Based On User Behavior

In the context of the development of the social applications and self-media,Zhihu has gradually become an information carrier and transmission medium that connects elites in various industries,discusses social public topics and shares knowledge and experience.As the Zhihu user community grows stronger,discover the users who are aware of the key role in information dissemination,it plays an important role in ensuring the security of the network environment,reasonable control,and guiding the direction of network user's opinion.This article considers the various behaviors of users and designs a more comprehensive evaluation method.The main work can be considered from the following three aspects:(1)Firstly,qualitative analysis of various behavioral characteristics of Zhihu users.From the perspective of information dissemination theory,it is found that the user influence of the knowledge is mainly divided into:dominant influence and recessive influence.Through the membership degree analysis to construct the index system of user behavior factors,the AHP method is used to calculate the weight of indicators,and the key factors of three dominant influences are determined:user professionality,user activity and user's content acceptance,these three factors are used to determine the node's influence;the key factors that affect the user's implicit influence are constructed:the interaction strength,the fan activity,and the influence coefficient between followee and his follower.(2)Next,build Zhihu social network,take the number of follower as in-degree,take the number of followee as out-degree,research the network characteristics;build an informed user influence based on the independent cascading model.The propagation model uses the influence coefficient determined in(1)as the probability value of the influence propagation edge;the HITS algorithm is used to take user as the node,take the attention relationship between the users as the edge,and the node influence as the initial value selection factor.Based on the influence coefficient as the weight of the edge,build the user influence model.Based on the evaluation model,the user influence evaluation algorithm-ZHUR-UB(ZhiHu user rank based on user behavior)is used as a quantitative tool for user influence.(3)Finally,according to the experimental data,the network characteristic attributes such as in-degree out-degree and clustering coefficient are analyzed to verify the validity of the data;analyze the user's main behavior data,indicating that these behaviors are used as the rationality of impact analysis,comparing the behavior data of fan users and non-fan users,indicating that the rationality of the previous considerations;calculate in a typical subset of data,the high-influence user top 10 sorting of HITS algorithm and ZHUR-UB algorithm is given.As a result,the ranking results are compared with the top 10 users of the number of follower,and the number of users who have excellent answer title in the sorting result are compared,and the answers of the top 10 to top40 users obtained by the two algorithms are respectively compared.The comparison of the three indicators of answer number,thanks number,and follower number,it shows that the ZHUR-UB algorithm can reflect the user behavior in many aspects and verify the rationality and effectiveness of the algorithm.