Educational data mining has become an important research field in studying the social behavior of college students using massive data. However, traditional campus friendship network and their community detection algorithms, which lack time characteristics, have their limitations. This paper proposes a new approach to address these limitations by reconstructing the campus friendship network into weighted directed networks in continuous time, improving the effectiveness of traditional campus friendship network and the accuracy of community detection results. To achieve this, a new weighted directed community detection algorithm for campus friendship network in continuous time is proposed, and it is used to study the community detection of a university student. The results show that the weighted directed friendship network reconstructed in this paper can reveal the real friend relationships better than the initial undirected unauthorized friendship network. Furthermore, the community detection algorithm proposed in this paper obtains better community detection effects. After community detection, students in the same community exhibit similarities in consumption level, eating habits, and behavior regularity. This paper enriches the theoretical research of complex friendship network considering the characteristics of time, and also provides objective scientific guidance for the management of college students.

翻译：教育数据挖掘已成为利用海量数据研究大学生社会行为的重要研究领域。然而，传统校园友谊网络及其社区检测算法缺乏时间特征，存在局限性。本文提出了一种新方法，通过将校园友谊网络重构为连续时间下的加权有向网络，克服了传统校园友谊网络的有效性局限并提升了社区检测结果的准确性。为此，我们提出了一种适用于连续时间下校园友谊网络的新的加权有向社区检测算法，并将其应用于某高校学生的社区检测研究。结果表明，本文重构的加权有向友谊网络比初始的无向无权友谊网络更能揭示真实的友谊关系。此外，本文提出的社区检测算法取得了更好的社区检测效果。社区检测后，同一社区的学生在消费水平、饮食习惯和行为规律性方面表现出相似性。本研究丰富了考虑时间特征的复杂友谊网络理论，也为大学生管理提供了客观的科学指导。