Research On Multiplex Network Community Detecting By Deep Learning Method

There are community structures exist in multiplex networks.Although it has acquired some achievements in multiplex network community structure detection,traditional multiplex networks community structure detection methods have high complexity,they are not suitable for processing large scale and high dimension multiplex networks.Besides,most of the methods proposed for community detection have the defect that losing important information of the original system,this missing information is the importance of single networks,causing the decline of partition accuracy in multiplex network societies.On the basis of summarizing and analyzing the existing multiplex network community detection technology,this paper puts forward a new multiplex network community detection algorithm based on deep learning method:First,a feature compression algorithm based on multiplex network community relation is proposed.In view of the shortcomings of the current multiplex network community detection algorithms,the property of network correlation degree and network importance is used,through these two concepts we make up the deficiency of the existing algorithms.Based on the sparse characteristic of large-scale multiplex networks data,a new algorithm is proposed to compress the characteristics of multiplex network Association,at the same time,the relation feature reconstruction algorithm is described.The experimental results show that the algorithm can effectively extract the information of multiplex network structure.Second,a multiplex networks community detection method based on deep learning is proposed.Deep belief networks have high performance for processing large scale data sets,and the algorithm is suitable for the community detection of large scale and high dimension multiplex networks.Using deep belief networks to reduce the dimension of multiplex network data,and using community labeled samples for supervised tuning.Also,the method of obtaining the training representative set,unsupervised learning method and method of supervised optimization are described in detail.According to the algorithm proposed in this paper,the experimental simulation is carried out on the real multiplex network data sets in different sizes.The feasibility and effectiveness of the multiplex network community relation feature compression algorithm are verified,and the performance of the multiplex network community detection method is verified.Compared with the classical multiplex network community detection algorithm,the experiments show that the proposed algorithm is more efficient than the existing algorithms to deal with large scale multiplex networks.