| Community detection is one of the hot spots in the study of complex networks,which aims to find subgraphs that are densely connected internally but sparsely connected externally.Traditional community detection algorithms have been very successful to a certain extent.However,data are often collected from multiple sources.Traditional community identification methods based on community topology can no longer achieve the desired results when faced with such multi-view data collected from multiple perspectives.To address these issues,this work presents a multi-view integration method for semantic social networks and a multi-view community discovery method based on adaptive loss functions,which provides an effective representation for semantic features at different granularity levels and rationally exploits the diversity and complementarity of different views in social networks to achieve community segmentation.To sum up,the main work includes:1.Representation of semantic features from multiple perspectives.According to the text processing strength,the user semantic information in the social network is extracted from the three perspectives of word frequency,keyword s and topics.Complete the mapping of node semantic information to node feature vector,and realize the semantic feature representation of social networks.2.Reconfiguration methods for social networks.Taking advantage of the robustness of sparse representation to noise and outliers in data,a similarity matrix learning method based on sparse representation is constructed to establish connections between users with similar semantic information and realize the reconstruction of social network.3.Construction of multi-view community detection methods.A multi-view community detection method based on adaptive loss functions(ALMV)is proposed,which utilizes the robustness of the L21-norm and F-norm to construct an adaptive loss function that maximizes the mutual information of each view.Based on the above,an optimization expression is constructed to generate a unified graph matrix that outputs a multi-view community structure.4.Experimental results and analysis.Through simulation experiments,the algorithm is compared with eight baseline methods on real social network datasets and eight public datasets to verify the effectiveness and performance of the algorithm.According to experimental results,the method is effective on real social networks and outperforms other baseline methods in terms of stability and performance,yielding high-quality community structures. |
