Research And Application Of Non-overlapping Community Division Algorithm And Privacy Protection

In the real world,everything is connected with each other,and these connections can be abstracted into complex network models.The network can be divided into several communities according to the degree of relationship between entities in the complex network.This community structure is one of the important characteristics of the complex network.In order to analyze the community structure and understand the hidden information in the network,the community must be divided first,and before the community is divided,the corresponding network data set needs to be obtained.How to ensure the privacy and security of the data set and the accuracy and efficiency of the community division results are both current research hotspots.Based on the research of several classic non-overlapping community division algorithms and privacy protection algorithms,two improved community partition algorithms and a privacy protection method based on anonymity are proposed for the defects and deficiencies in those algorithms.The details are as follows:Label propagation algorithm IRA-LPA,which is based on improved RA similarity.Aiming at the situation that there are many related nodes in the calculation results of RA similarity algorithm,but the similarity is zero or cannot be calculated,IRA-LPA algorithm introduces Sim Rank algorithm to improve RA similarity,and then combines the improved RA similarity with the label propagation algorithm to improve the accuracy and performance of the label propagation algorithm.Community division algorithm JLCD,which is based on Jaccard and LPA.Aiming at the shortcomings of the Jaccard similarity algorithm,the algorithm corrects the unreasonable values in the Jaccard similarity matrix,and then applies the results of improved Jaccard algorithm to the LPA algorithm for final community division,which reduces the resources consumption of label initialization in the label propagation algorithm.Social network privacy protection method NI-PP,which is based on the importance of nodes.This algorithm introduces the concept of node importance into anonymity protection,combines k-symmetric anonymity technology to achieve the purpose of focusing on protection of important nodes,and provides a restoration algorithm that can completely restore the original network structure to improve the usability of the anonymous network and reducing information loss in network restoration.The algorithms and methods proposed in this thesis have been applied to the classical data sets and artificially generated data sets for testing.The experimental results show that the three algorithms have good effects in different application scenario,IRA-LPA and JLCD can effectively divide the community and have higher accuracy and lower time complexity and NI-PP can provide higher confidentiality and usability.