|
The development of location service technology provides the premise for the emergence of location-based social network (LBSN).At present, the research on community discovery and social relations in LBSN still has some problems, such as insufficient model dimension, difficulty in parameter estimation and low accuracy of model results, which are still challenges for researchers.In order to solve the problem of poor model fitting effect, this paper designs a multi-dimensional social relationship analysis model (MSRAM) from two aspects of social relationship and user check-in behavior by introducing dynamic interaction relationship between nodes.Then, according to the hidden variable sampling rules and parameter iteration rules, a new Gibbs sampling algorithm based on model is proposed, and the experimental parameters are estimated according to the cross-entropy reality.By comparing with the spectral clustering algorithm and the TGSC-PMF algorithm based on the topic model, it is verified that the cross-entropy of the proposed model is increased by about 7.93%, the modularity is increased by 5.6%-8.8%, and the similarity within the community is increased by 12%-23%.The results show that this model can achieve more accurate community division for the whole experimental data network. |
|
Keywords:Technology of Computer Application; social network; LDA topic model; dynamic interaction |
|