面向个性化课辅的学生学习行为画像及其应用研究

To achieve the individualized learning, the applications of the "Internet + Education" have sprung up constantly. However, how to depict students' learning behaviors and how to improve their learning efficiency and learning effect are significant research topics based on their learning behaviors. In the one hand, based on the metacognition theory and the cognitive diagnosis model, we aim at profiling the learning behaviors of students in the knowledge layer and the learning quality layer via utilizing association rules, frequent item mining, matrix factorization and subgraph matching. In the other hand, we plan to study the accurate individualized tutoring by using the methods of probabilistic matrix factorization, set covering, deep knowledge tracing and real-time optimization of the particle swarm algorithm, etc. To provide data and platform supports for the related researches of the learning behaviors profiling and its applications, we are going to design and implement a prototype system and application demonstration for the learning behavior profiling and its applications toward the individualized course tutoring. Solving these problems is helpful to the applications of Internet + Education, such as electronic mistake collection, individualized tutoring, learning ability evaluation and educational decision making, etc. It starts from the knowledge layer and the learning quality layer profiling, and aims at achieving information transformation of "quantity-quality-order". It therefore makes an innovation in the technical route. The applicant has studied the research problem of user profiling and its applications for a long time, and has achieved many research results. The project is therefore clear in objectives and strong in feasibility.

为实现因材施教，“互联网+教育”的应用如雨后春笋般不断涌现。然而，如何刻画学生学习行为以及如何应用学生学习行为提高学习效率、提升学习效果成为一个重要的研究课题。本课题基于元学习理论和认知诊断模型，运用关联规则、频繁项挖掘、矩阵分解和子图匹配等方法寻求学生学习的知识层和品质层画像；运用概率矩阵分解、集合覆盖、深度知识寻径和粒子群实时优化等方法探索精准的个性化辅导；在此基础上设计面向个性化课辅的学生学习行为画像及其应用的原型系统。解决这些问题有助于电子错题本、个性化辅导、学习能力评价和辅助决策等互联网+教育的应用。课题以细粒度的学生学习知识层和学习品质层画像为切入点，实现学生学习行为数据的“量-质-序”的转化，在技术路线上具有新意。申请人在用户画像及其应用等方面有长期丰富的积累，具有顺利开展课题研究的良好基础，课题目标明确，可行性强。

为实现因材施教，“互联网+教育”的应用如雨后春笋般不断涌现。然而，如何刻画学生学习行为以及如何应用学生学习行为提高学习效率、提升学习效果成为一个重要的研究课题。本课题基于元学习理论和认知诊断模型，运用关联规则、频繁项挖掘、矩阵分解和子图匹配等方法寻求学生学习的知识层和品质层画像；运用概率矩阵分解、集合覆盖、深度知识寻径和粒子群实时优化等方法探索精准的个性化辅导；在此基础上设计面向个性化课辅的学生学习行为画像及其应用的原型系统。解决这些问题有助于电子错题本、个性化辅导、学习能力评价和辅助决策等互联网+教育的应用。在此项目资助下，共计发表论文19篇，其中CCF-A类国际会议和期刊论文6篇，CCF-B类国际会议和期刊论文6篇，培养毕业博士研究生1人、硕士研究生8人，获得国家科技进步二等奖1次、上海市科技进步一等奖1次、上海市教学成果一等奖1次。