Research And Application Of Hybrid Overlapping Community Detection

Community detection has developed rapidly in recent years as an important method of complex network research. Detect communities in complex network can not only help people understand the organizational structure of the network, but also can dig out a lot of hidden information. This information can be used in data organization, e-commerce, personalized recommendations and other practical applications. It is considerable difficult to detect communities in large-scale complex network rapidly and accurately.The coauthor network is a kind of social network. We can get the author's domain division results by means of detecting community in the coauthor network. However, it is also worthy to discuss how to ensure those results has practical significance. For the above problems, this paper mainly studies the following aspects:(1) This paper introduces the situation of the community detection study firstly, then analyses and compares the advantages and disadvantages of several existing community discovery algorithms(2) Here puts forward a local community detection algorithm based on the community contribution coefficient. The algorithm is suitable for large-scale network. Compare with the existing local community discovery algorithm, it improves the accuracy, overcomes the weakness of sensitivity of the initial node selection and avoids the problem of the resolution limitations.(3) Here then proposes a hybrid overlapping community discovery algorithm. The algorithm consists of two parts: node splitting and community detection. The improved process of node splitting reduces the time complexity, and combines with the previous proposed local community Discovery algorithm to achieve the overlapping community detection.(4) The application of hybrid overlapping community discovery algorithm in the author's domain division is proposed. The previous proposed algorithm is used in the coauthor network constructed according to the DBLP data set. It can detects all of the author's domain which the given author belongs to and attains the goal of domain division.This paper also designs experiment to implement the algorithms on several different data sets. The analysis of the experimental results proves the effectiveness of the algorithms. Finally, the application of the algorithms in author's domain division shows that the algorithms have practical meaning.