关键词: 重叠结构, 高校社团, k-派系过滤算法, 复杂网络, 超网络

Abstract: In most real-world networks, there are no completely independent community structures, and they are usually made up of many intertwined and overlapping communities. An overlapping community structure refers to nodes that can belong to multiple different communities at the same time. For example, in a scientific collaboration network, some scientists may be both biologists and mathematicians. If we classify this network according to different disciplines, the same individual may be assigned to two different communities. In university student associations, some students participate in multiple associations, so as to be categorized into various communities. University student associations have the following characteristics: First, there are many organizations with a wide variety and large number of participants. Second, associations are becoming increasingly diverse. Different students' interests and hobbies are quite different, forming a wide range of university associations. Third, each organization establishes its own rules and regulations, with a more systematic management approach. The overlapping structures within university student associations represent the cross-penetration between communities. Studying the cross-penetration of university organizations from the perspective of the overlapping structure is a worthwhile topic to explore.
Based on this, this paper collects a total of 6,580 data points on university students' participation in associations through surveys, QR code scanning, and hyperlinks. It removes responses from individuals who do not participate in any associations and those who have no common associations with any classmates, resulting in 5,040 valid data points. Each university student is treated as a node, and since students within the same organization know each other, a fully connected network of nodes belonging to the same organization is formed, creating edges for the network and constructing the university organization network. Using the k-clique percolation algorithm, we perform the overlapping structure detection and community partitioning. Furthermore, based on different values of k, we analyze overlapping relationships within the network by combining certain metrics of complex networks and hypernetworks.
By comparing the ratio of retained nodes, number of community partition, and modularity Q value across different values of k to the actual partition results from the empirical dataset, the effectiveness of the algorithm is validated, leading to the optimal value of k. This paper helps to analyze the structure and characteristics of university associations, further puts forward guiding suggestions for the construction of associations in three aspects, namely, attaching importance to the guiding role of schools, strengthening the construction of association talents and association culture, and avoiding blind obedience in joining associations, which provide a theoretical basis for the construction and service of university associations and the selection of associations by college students, and also has certain practical significance.
In future research, we will combine the characteristics of overlapping communities to find a fast and reliable community detection algorithm, and focus on the community detection method for hypergraph to ensure the practicability of the method. In addition, this paper studies the static undirected unweighted network, but due to the network evolution of new nodes and the new relationship, makes the network as a whole in a dynamic change. There are some directional edges in the real network, so there are a lot of weighted and directed networks. How to extend the community detection method to these networks is also the next research direction.

Key words: overlapping structure, university associations, k-clique percolation algorithm, complex network, hypernetwork