Due to internal disputes, the club splits into two groups, which is its real network community structure. NCAA College-Football Network. The network of American football games between Division IA colleges during TAK-700 molecular weight Regular Season Fall 2000 (http://networkdata.ics.uci.edu/data.php?id=5) is composed of 115 vertexes and 1,232 edges, in which each vertex corresponds to an American college football team and each edge represents two corresponding teams played a game during Regular Season Fall 2000. All the teams are divided into eleven conferences and five independent teams. Books about US Politics. The network of books about recent US Politics sold by the online bookseller

is composed of 105 vertexes and 882 edges, in which each vertex corresponds to an US Politics book and each edge

represents the frequent copurchasing of two corresponding books. DBLP Coauthorship Network. A weighted network of authorship in four research fields (i.e., DB, IR, DM, and ML) extracted from the DBLP computer science bibliographical dataset is composed of 28,702 vertexes and 66,832 edges, in which each vertex corresponds to a distinct author who has published more than twenty papers and each edge represents their coauthor relationship. The weight of an edge denotes the number of papers coauthored by these two authors. Meanwhile, we utilize the tool developed by Lancichinetti et al. [17] to generate several synthetic networks and divide them into two groups based upon the number of nodes in networks, with the nodes number of one group being 1000 and the other group 10000. Each group comprises 15 networks, with their mixing coefficient ranging from 0.1 to 0.8 at a step size of 0.05. To further evaluate the performance of our method, we also run our algorithm on networks

of different number of nodes, including 1000, 5000, 25000, 5000, 100000, 250000, and 500000, with the mixing coefficient being 0.3. 4.2. Analysis of the Influence of Parameter α To compare the impacts of different values AV-951 of α on the performance of our algorithm, we conduct our experiment on the benchmark Football dataset and fifteen 1000-node synthetic LFR networks with their mixing coefficients varying from 0.1 to 0.8 at an increment interval of 0.05. Setting the values of α from 1 to 40, when detecting communities in the real network Football and the synthetic networks, the NMI values of our algorithm are shown in Figures 4(a) and 4(b). Figure 4 The achieved NMI values of our algorithm varying with the parameter α in a real network Football and the synthetic networks with n = 1000. As shown in Figure 4(a), in the real Football network, when α = 2, the highest NMI value is obtained, indicating that the results are the closest to the correct ones.