大规模在线课程中用户流失问题的研究

项目名称： 大规模在线课程中用户流失问题的研究

项目编号： No.61472006

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 自动化技术、计算机技术

项目作者： 张铭

作者单位： 北京大学

项目金额： 83万元

中文摘要： 大规模在线公开课（简称MOOC）自2012年起飞速发展，使得全世界范围内的学生可以免费聆听世界一流高校课程内容。以Coursera为代表的MOOC平台，吸引了超过600万的注册用户，并有600多门课程。MOOC在工业界和学术界都受到广泛关注的同时，较高的用户流失率成为了亟待关注和解决的问题。研究用户流失问题在各方面都有重要意义，例如提高MOOC平台的利用率、改善教学设计、及时给予用户激励措施等等。本课题从用户的固有静态属性、个人动态行为、社区行为三个方面进行研究；并以统计分析、机器学习技术、概率图模型、深度学习、迁移学习以及教育学理论为工具，对用户的流失问题进行全方位的研究；再将用户学习行为与流失率联系起来进行分析与建模，预测用户是否会流失，并深入分析用户流失的具体原因，以提高MOOC资源的社会效益。

中文关键词： 大规模公开在线课程；用户流失；用户静态属性建模；用户动态行为建模；用户社区行为建模

英文摘要： Massive Open Online Course (MOOC) has been witnessing an impressive growth rate since 2012, which enables students from all over the world to get free access to the courses from world-class universities. Coursera, one of the most successful MOOC platforms, has attracted more than 6 million registered users and more than 600 courses. While MOOC is getting more and more attention in industry and the Academic community, the low rate of user retention is becoming an important and urgent problem. Understanding users' disengagement behavior is very important and can possibly support many applications, such as increasing usability of MOOC platform, improving teaching design, motivating students at the right time. We focus on analyzing the user retention problem by modeling users from three aspects, which are modeling users' static attributes, modeling users' dynamic study behaviors and modeling users' behaviors in the learning community. By doing that, we apply the knowledge of statistical analysis, machine learning, probabilistic graphical models, intensive learning, transfer learning and pedagogy. We also propose a model to predict users' retention behavior and analyze the reasons for disengagement so as to improve the social benefit of MOOC resources.

英文关键词： Massive Open Online Course (MOOC);User retention;Modeling Users’ Static Attribute;Modeling Users’ Dynamic Study Behavior;Modeling Users’ Community Behavior