睡眠障碍是重要的精神健康问题。近年来，大学生群体的睡眠障碍问题益发严重，是导致其焦虑、抑郁等精神疾病甚至自杀行为的重要因素。睡眠障碍的诊断与相关研究对于公众心理健康发展至关重要。围绕应激事件诱发个体睡眠障碍这一心理学问题，本项目原创性地提出基于微博平台对大学生群体的慢性睡眠障碍进行诊断并对其心理轨迹进行溯源，以实现对高危睡眠障碍个体的甄别和理解。在慢性睡眠障碍的诊断上，针对个体微博数据稀疏、睡眠节律多变的特点，提出通过构建基于发博行为的概率模型，对慢性睡眠障碍区间及睡眠节律变化模式进行检测，解决单点检测中可能存在的漏检误检问题；在诱因事件的检测上，拟基于领域专家知识构建语料库，结合深度学习框架和迁移学习策略，对应激事件细节进行抽取，在此基础上给出关键应激事件的归纳策略。本项目回应了传统心理学研究难以触及的大规模、及时性等难题，为慢性睡眠障碍的诊断提供了全新的解决思路。

Sleep disorders are important mental health issues. In recent years, the problem of sleep disorders among college students has become more and more serious, which is an important factor leading to their anxiety, depression and other mental illnesses and even suicidal behaviors. The diagnosis of sleep disorders and related research are crucial to the healthy development of our society. Focusing on the psychological problem of individual sleep disorders induced by stressful events, this project originally proposed to diagnose the chronic sleep disorders of college students based on microblog platform, so as to realize the identification and understanding of individuals with high-risk sleep disorders. In the diagnosis of chronic sleep disorders, according to the characteristics of sparse individual microblog data and changeable sleep rhythm, this project proposed to construct a probabilistic model based on microblogging behavior to detect the chronic sleep disorder interval and sleep time change patterns. In the detection of incentive stressors, this project proposed to build a corpus based on domain expert knowledge. By combining the deep learning framework and the transfer learning strategy, the details of stressful events can be extracted. On this basis, the induction strategy of key stressful events was proposed. This project responds to the problems of large-scale and timeliness that are difficult to be reached by traditional psychology research, and provides a new solution to the diagnosis of chronic sleep disorders.