时间模式可以呈现学习行为随时间变化的规律，具有重要的研究与应用价值。相关研究主要局限在学习资源访问的“热点”现象、学生点击流的“潮汐”规律等统计摘要方面，难以呈现细粒度的时间模式。本项目针对多任务和时间碎片化等学习特点，以揭示学习行为随时间变化的规律为目标，拟开展多维学生情境模型构建研究、基于协同感知的时间模式提取研究和时间知识的智慧呈现研究。首先，通过监督和非监督学习将数据划分成簇内同质、簇间异质来建立协同感知环境；其次，从多变量时间序列中获得兴趣度模型实现对序列特征的重要性采样，然后将采样序列放到对应簇群中进行协同训练，从而迭代地更新模型和进一步优化序列选择；再次，通过研究时间图谱构建等技术实现时间知识的表示和结构化查询；最后，研制可视化原型系统并在大规模e-learning中验证。通过本研究，能细粒度呈现大时间窗口中学习行为随时间变化的规律，对在线学习的个性化诊断和评价具有重要意义。

The temporal pattern can be used to present the rules of learning behaviors changing over time, having significant research and application values. However, relevant studies with respect to temporal pattern are mainly limited to statistical abstracts regarding the accessing prevalence of learning resources and the tidal regularities of user click streams etc., which are difficult to present fine-grained temporal patterns. In view of learning has characteristics of multitasking and time fragmentation etc., this project aims to discover the rules of learning behaviors changing over time through studies of modeling multi-dimensional student contexts, extracting time patterns based on collaborative awareness, and smart presentation of temporal knowledge. Firstly, the supervised and unsupervised learning is used to divided the data into homogeneity within clusters and heterogeneity among clusters to establish the collaborative awareness evironment. Secondly, a sequential model with the interest measure is obtained from multivariate time series to make importance sampling to the sequential features, and then the sampled sequences are collaboratively trained in corresponding clusters in order to update the model iteratively and further optimize the sequence selection. Thirdly, the temporal graph construction and other techniques are used to realize the presentation of temporal knowledge as well as its structured query. And finally, a visualized prototype system is developed and tested in a large-scale e-learning scenario. Though this project, the patterns of learning behaviors changing over time hidden in big time windows are discovered in more detail, which is of great significance to the personalized diagnosis and evaluation of online learning.