华东师范大学学报(教育科学版) ›› 2022, Vol. 40 ›› Issue (11): 110-122.doi: 10.16382/j.cnki.1000-5560.2022.11.009

• 学习科学与研究方法 • 上一篇    下一篇

贝叶斯网络方法能为教育研究带来什么?

顾昕1, 毛梦琪1, 马淑风1, 陈森宇2   

  1. 1. 华东师范大学教育学部教育心理学系,上海 200062
    2. 曼彻斯特大学健康科学学院,曼彻斯特 M13 9PL
  • 出版日期:2022-11-01 发布日期:2022-10-27
  • 基金资助:
    上海市浦江人才项目“面向核心素养测评的贝叶斯网络算法开发与优化策略”(2020PJC034);华东师范大学幸福之花项目“复杂学习情境下核心素养测评范式及其培养机制研究”(2019ECNU-XFZH015)

What Can Bayesian Network Contribute to Educational Research?

Xin Gu1, Mengqi Mao1, Shufeng Ma1, Senyu Chen2   

  1. 1. Department of Educational Psychology, Faculty of Education, East China Normal University, Shanghai 200062, China
    2. School of Health Sciences, University of Manchester, Manchester, M13 9PL, UK
  • Online:2022-11-01 Published:2022-10-27

摘要:

针对当前教育研究问题的复杂性、不确定性与动态性特征,本文提出使用贝叶斯网络方法分析教育实证研究数据。在研究范式上,贝叶斯网络结合了理论驱动与数据驱动的研究方法,根据教育研究理论与专家经验确定先验模型,通过后续采集的数据迭代模型,不断更新能够支持或反对理论模型的数据证据。在数据分析方法上,贝叶斯网络将变量或变量关系的不确定性纳入模型,以概率的方式给出精确的推断结论和预测信息。在模型应用上,贝叶斯网络能够在真实教学情境中实时评估学生的知识掌握、能力培养、素养发展等,为教学与学习过程的动态评估提供方法和技术上的支持。

关键词: 贝叶斯网络, 概率推理, 先验知识, 教育过程数据

Abstract:

This paper proposes to use the Bayesian network method to analyze educational research data in view of the complexity, uncertainty and dynamic characteristics of current educational research problems. In terms of research paradigm, Bayesian network integrates theoretical and data-driven research procedures, determines a prior model based on educational research theories and expert experience, and updates data evidence that can support or oppose the theoretical model by collecting new data. In terms of data analysis, Bayesian network brings the uncertainty of variables or variables’ relations into the model and gives accurate inference and prediction by means of probability. In terms of application, Bayesian network can evaluate students’ knowledge mastery, ability training and competence development in real teaching and learning situations. This offers methods and technical support for dynamic evaluation of teaching and learning process.

Key words: Bayesian network, probabilistic inference, prior knowledge, process data in education