Method for user influence evaluation in social network

Abstract

The invention relates to a method for user influence evaluation in a social network. The method comprises the following steps of step A: reading social network data, and configuring a social network diagram G with social network users as nodes and user relationships as sides; step B: traversing all the nodes in the social network diagram according to the social network diagram, initializing an influence label of each node according to the degree of the nodes, and finishing traversing; step C: traversing all the nodes in the social network diagram according to the social network diagram, and computing the influence levels of the traversed nodes according to the influence levels of the neighbor nodes of the traversed nodes; step D: repeating the step C until the influence level of each node is converged. The method has the advantages of linear time complexity which is approximately linear and capabilities of effectively analyzing the user influence distribution situation in the large-scale social network and excavating high-influence users, and can be applied to the fields, such as network marketing.

Description

Technical field [0001] The invention relates to the technical field of social network analysis, in particular to a method for evaluating user influence in social networks. Background technique [0002] Social influence refers to a phenomenon in which users, organizations, or communities have social relationships with other users, organizations, or communities, which causes their behavior to change with changes in other users, organizations, or communities. Social influence is a common phenomenon in social networks. In social networks, a variety of factors may have an impact on influence. By analyzing the influence of nodes in social networks, it is possible to find core nodes with important influence in social networks, which can be used in many fields such as corporate commercial marketing, targeted advertising, speech channel recommendation, and public opinion monitoring. [0003] At present, the influence analysis methods on nodes mainly include two categories. One method is b...

AI Technical Summary

Problems solved by technology

Using the degree to evaluate the influence of nodes is suitable for non-uniform networks that conform to the power law, but it ignores the rich topological characteristics of the network; the betweenness is the number of shortest paths passing through a node to describe the importance of nodes , although the topology of the network is taken into consideration, its complexity is too high to be suitable for large-scale real networks; the results obtained by K-shell decomposition have a large number of k - Nuclear nodes, which do not match the diversity of nodes in social networks

Based on the idea of ​​K-shell decomposition, Zeng et al. proposed the MDD (Mixed Degree Decomposition) method to distinguish between k - Kernel value for node influence, but the method's parameter lambda It is difficult to determine the optimal value in different networks; another method is the web page ranking algorithm based on link analysis, such as the classic PageRank and HITS methods and their improved methods; such as Zhu Tian et al. Based on the community structure of social networks, proposed InnerPageRank and OutterPageRank are two evaluation methods, which are used to calculate the influence of nodes inside and outside the community respectively

Such methods require repeated iterations, high time complexity, and weak general applicability

[0004] In summary, the existing influence assessment methods for individual users in social networks are difficult to meet the requirements in terms of analysis effect and efficiency in the face of large-scale social network scenarios.

Images