Abstract: With the advancement of eye tracking technology and the reduction of equipment costs, eye tracking technology has been widely applied in the field of intelligent education. Therefore, analyzing eye tracking data to evaluate learning status has become a very important part of intelligent education. Eye movement scanpath can reflect thinking pattern and mental state changes directly or indirectly. In this paper, scanpath is analyzed to explore the commonalities and differences of learners’ eye movement behaviors, so as to provide reference for improving visual stimulus and giving guidance. The time series representation and clustering of learners’ scanpath under the same task are studied, and the learning states are evaluated by clustering results, which are divided into three states：concentration, mind-wandering and information wandering. Based on the improved DBA（DTW barycenter averaging） algorithm to extract group eye movement patterns, combined with the DTW（dynamic time warping） algorithm to calculate the similarity and clustering seed of scanpaths, DDC （distance density clustering） algorithm is used to cluster the scanpaths. Experimental results show that eye movement pattern mining based on time series can identify group viewing behavior. Clustering reveals different reading strategies and provides the ability to assess learning status.

Key words: scanpath, eye movement pattern, learning?state, DTW barycenter averaging（DBA）, distance density clustering（DDC）

摘要： 随着眼动追踪技术的进步和设备成本的降低，眼动追踪技术已广泛应用于智能教育领域，分析眼动数据以评估学习状态成为智能教育中一个十分重要的环节。眼动扫描路径可以直接或间接地反映思维模式及心理状态的变化，通过分析扫描路径探索学习者眼动行为的共性和差异性，为改善视觉内容和给出指导性意见提供重要参考。首先研究在同一任务情况下学习者扫描路径的时间序列表示和聚类，通过聚类结果评估专注、走神及信息迷航等三种学习状态。进而对重心平均动态时间规整（DTW?barycenter averaging，DBA）算法进行改进，并用于提取群体眼动模式，结合动态时间规整（dynamic time warping，DTW）算法计算扫描路径的相似度和确定聚类种子，采用距离密度聚类（distance density clustering，DDC）算法进行聚类。实验表明，基于时间序列的眼动模式挖掘能够识别群体观看行为。而聚类揭示了不同的阅读策略，并提供了评估学习状态的能力。

关键词: 扫描路径, 眼动模式, 学习状态, 重心平均动态时间规整（DBA）, 距离密度聚类（DDC）