Overlapping community detection in complex networks such as social, biological, economic, and other real-world networks has become an important area of research in the past two decades. Community detection problem can be modeled as a multiobjective optimization problem and it has been successfully solved using Evolutionary Algorithms (EA). The representation scheme of a solution in an EA affects the size of search space and thereby its convergence. A clique-based representation scheme is suitable to represent overlapping communities since overlapping cliques represent overlapping communities. The community to which a clique belongs becomes the community of its containing nodes. However, there are two issues viz., (1) cliques may heavily overlap since they might share many nodes with other cliques (2) one-node and two-node cliques increase the length of the representation of the EA. This paper proposes a merged-maximal-clique based representation scheme which reduces the chromosome length with fewer number of cliques. The proposed scheme overcomes the drawbacks of the existing clique-based scheme and naturally brings the initial solutions of the EA closer to the actual solutions. Moreover, operators of the EA are enhanced by renumbering the solutions to avoid duplicate solutions taking part in evolution. An EA based on decomposition equipped with the new representation scheme and enhanced operators is tested on both real-world and synthetic networks. The proposed EA finds better community partitions in fewer generations compared to the existing clique-based and non-clique-based algorithms. This study proves that clique-based schemes are efficient for overlapping community detection.

在社会、生物、经济和其他现实世界网络等复杂网络中，重叠社区检测已成为过去二十年的重要研究领域。社区检测问题可以建模为多目标优化问题，并且已使用进化算法 (EA) 成功解决。EA 中解的表示方案会影响搜索空间的大小，从而影响其收敛性。基于团的表示方案适用于表示重叠社区，因为重叠团代表重叠社区。一个集团所属的社区成为其包含节点的社区。但是，有两个问题，即 (1) 集团可能会严重重叠，因为它们可能与其他集团共享许多节点 (2) 单节点和双节点集团增加了 EA 表示的长度。本文提出了一种基于合并最大集团的表示方案，该方案以较少的集团数量减少了染色体长度。所提出的方案克服了现有基于集团的方案的缺点，自然地使 EA 的初始解决方案更接近实际解决方案。此外，通过对解决方案重新编号以避免重复解决方案参与演化，EA 的运营商得到了增强。基于分解的 EA 配备了新的表示方案和增强的算子，在现实世界和合成网络上进行了测试。与现有的基于派系和非基于派系的算法相比，所提出的 EA 在更少的世代中找到了更好的社区划分。这项研究证明了基于集团的方案对于重叠社区检测是有效的。