基于拓扑信息和特征属性的社区发现算法

Community Discovery Algorithm Based on Topological Information and Feature Attributes

文章提出一种新颖的加权算法,旨在提高社区检测算法准确性。该算法结合网络拓扑信息和节点属性信息,采用光谱聚类技术来揭示潜在的社区结构。它能够自适应地学习不同属性对检测任务的贡献度并相应地分配权重,同时采用软阈值算子和L1正则化项来降低“噪声”属性对检测精度的不利影响。通过对数值模拟和真实世界数据集的测试验证该算法在社区检测应用中的有效性。

A novel weighted algorithm aimed at improving the accuracy of community detection algorithms is proposed.The algorithm integrates network topology information and node attribute information,utilizing spec⁃tral clustering techniques to reveal potential community structures.It is capable of adaptively learning the contribution of different attributes to the detection task and accordingly assigning weights,while employing soft threshold operators and L1 regularization terms to reduce the adverse impact of“noise”attributes on de⁃tection precision.The effectiveness of the algorithm in community detection applications has been validated through the tests on numerical simulations and real-world datasets.