Research On Dynamic Model Of User Behavior On Topic

With the rapid development of the internet and mobile internet,social network is increasingly becoming an important platform for people's daily communication and information dissemination.Therefore,the social network analysis technology has become an important way of user behavior analysis,public sentiment awareness and monitoring.As the user is the main body of social network,research on its relationship network and behavior rules is very important.Focusing on user behavior and relationship network,this thesis studies the influence factors of user's participation behavior and the discriminating of implicit relationship betwee users by model,algorithm and experiment.The specific works are as follows:1.In the aspect of user behavior analysis,this thesis mainly mines the driving factors and driving strength of user behavior by analyzing user profile and hostorical behavior.Firstly,experimental data is obtained from social network platform,including user profile and historical behavior.Secondly,taking the internal and external driving factors into account,personal interests,explicit links,implicit links and random driven attributes are extracted to formally define the driving factors,including personal driving factor,explicit driving factor,implicit driving factor and random driving factor.Finally,this thesis builds the user behavior dynamics model according maximum likelihood estimation theory to get the difference of the factors.The experimental result confirms that our model is able not only to confirm and quanlify the driving factors of user bebavior,but also to veritify the influence of implicit link.2.In the aspect of user relationship analysis,this thesis mainly discriminant the implicit relationship between users by analyzing their interests and behavior trace.Firstly,the topic information is extracted from social network,including the topic label,time and number of participating in the topic.Secondly,according to the idea of supervised learning,the dataset is divided and it mainly includes the explicit link users and non explicit link users of target user.Then,the optimal classification threshold is obtained to build the implicit link discriminant model by claculating the behavior similarity between target user and training set.Finally,the relationship between target user and test set users are discriminated.The experiment result shows that the model can not only discriminant the exciplit link relationship between users,but also mine the implicit link relationship between users.