Comparative Study on A Class of Evaluation Indices for Community Detection
Community detection and network partition are fundamental for uncovering the links between structure and function in complex networks. Recently Li et al. 11 introduced a novel quantitative function(D-value) for community detection which can overcome some drawbacks of the widely used modularity Q. We notice that although the modularity density Dλ has gained good performance for some networks, but how to determine a proper value of λ for any new network to be partitioned remains an open problem. This will certainly limit its further applications in practice to some extent. In this study, we. propose a general form G of evaluation index for community detection from a perspective of intuition, and its three typical forms are given. The simplest one is the linear form GL, which is just the D-value 11, the other two are the quadratic form GQ and the entropy function form(or logarithmic form) GE, respectively. By comparing the computational results on partitioning several real-world networks into communities we can conclude that GQ is inefficient, but GE is more powerful than GL(i.e., D in 11) to some extent. Moreover, the GE can also overcome some drawbacks of Q, and it doesnt contain any parameters, so it is very convenient for using in practice.
Community detection modularity complex networks
Junhua Zhang Shihua Zhang Xiang-Sun Zhang
Academy of Mathematics and Systems Science, Chinese Academy of Sciences,Beijing 100190, China
国际会议
The Second International Symposium(OSB08)(第二届国际优化及系统生物学学术会议)
云南丽江
英文
294-303
2008-10-31（万方平台首次上网日期，不代表论文的发表时间）