Learning Analytics to Support Teachers’ Assessment of Problem Solving: A Novel Application for Machine Learning and Graph Algorithms
Learning Analytics to Support Teachers’ Assessment of Problem Solving: A Novel Application for Machine Learning and Graph Algorithms
复制标题
支持教师评估问题解决能力的学习分析：机器学习和图算法的新颖应用
DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
V. K. Gupta
中科院分区：
文献类型：
--
作者：
P. Giabbanelli;A. Tawfik;V. K. Gupta
与具有预定义正确答案的结构良好的问题相比，复杂的现实世界问题通常是结构不良的问题（ISP）。ISP的开放性给评估和指导学生形成更好的解决方案造成了相当大的障碍，导致探究式学习的采用率很低。学生可以以知识图或因果图的形式为ISP构建和表示他们的知识，这些知识图阐述了相关概念及其因果关系（即，前因和后因）。评估这样的图可以涉及无参考评估（例如，以鼓励创建具有高密度概念的地图）或与用作参考的专家地图进行比较。本章首先回顾了将学生地图与专家地图进行比较的理论和工具。以前的方法通常比较单个连接（例如，与专家相比对学生具有/错过的连接的数量进行评分）或一般地图度量（例如，一个图比另一个图更密集）。相比之下，比较两个地图的问题已经在网络理论和图论中研究了几十年，产生了目前在教育研究中未充分利用的算法类别。本章回顾了三类算法（即，图核、图编辑距离、图嵌入），并将其应用于评估和学生成功。我们讨论了这些算法的实现，通过一套新的数字化工具，旨在支持社区的实践，以问题为基础的教学。
In contrast to well-structured problems which have pre-defined, correct answers, complex real-world problems are often ill-structured problems (ISPs). The open-ended nature of ISPs creates considerable barriers to assess and guide students in forming better solutions, which results in low adoption levels of inquiry-based learning. Students can structure and represent their knowledge for an ISP in the form of knowledge maps or causal maps, which articulate relevant concepts and their causal relations (i.e., antecedents and consequents). Assessing such maps can involve a referent-free evaluation (e.g., to encourage the creation of maps with high density of concepts) or a comparison to an expert map used as reference. This chapter starts with a review of theories and tools to compare a student’s map to the expert map. Previous approaches often compared individual connections (e.g., scoring the number of connections that a student has/misses in contrast with the expert) or general map metrics (e.g., one map is denser than the other). In contrast, the problem of comparing two maps has been studied in network theory and graph theory for several decades, yielding categories of algorithms that are currently underutilized in educational research. This chapter reviews three categories of algorithms (i.e., graph kernel, graph editing distance, graph embedding) in light of their application to assessment and student success. We discuss an implementation of these algorithms through a new set of digital tools, designed to support a community of practice in problem-based instruction.
从对话中进行论证挖掘
DOI：
--
发表时间：
2014
期刊：
--
影响因子：
--
作者：
Budzynska K
通讯作者：
Budzynska K