Abstract: Mining university students’ social ties based on their spatio-temporal trajectory is one of the hotspots of educational big data research, some mining methods have also been proposed. In view of the impact of spatio-temporal distribution aggregation, the contribution of spatio-temporal factors to the formation of social ties is comprehended, the location weight and group weight based on Shannon entropy are constructed. The method for inferring Weighted Social Ties based on Multiple Hypothesis Test（WST-MHT） is proposed and used to explore the social ties and intimacy of university students. A lot of experiments have been done with real data sets. The result is compared with the current popular methods of inferring social ties. It is observed that the accuracy by using the WST-MHT is 98.9% at the optimal threshold, and the recall is increased by around 20%.

Key words: spatio-temporal aggregation, location weight, group weight, multiple hypothesis test

摘要： 基于时空轨迹数据挖掘大学生社交关系是教育大数据研究的热点之一，也提出了一些挖掘方法。针对时空分布集聚性的影响，综合考虑时空因素对社交关系形成的贡献度，并构建基于香农熵的地点权重和事件分组权重。提出一种基于多重假设验证的加权社交关系推断方法（WST-MHT），并运用于挖掘大学生的社交关系以及亲密程度。用真实数据集进行大量的实验。将结果与当前流行的推断社交关系的方法进行比较，观察到在最优阈值下，WST-MHT的准确度达到98.9%，并且召回率提升了约20%。

关键词: 时空集聚性, 地点权重, 分组权重, 多重假设验证