中国教育类核心期刊 CSSCI来源期刊 RCCSE中国权威学术期刊

杂志文章

您的位置:首页 >文章目录 >2018年第10期>

    多知识点知识追踪模型与可视化研究

    Research on Knowledge Tracing Model for Multiple Knowledge Points and Visualization

    [浏览次数:7773]

【作      者】:

徐墨客, 吴文峻, 周 萱, 蒲彦均


【关 键 词 】:

智慧教育; 学生能力评估; 可视化; 知识追踪模型


【栏      目】:

学习环境与资源


【中文摘要】:

知识追踪模型(简称KT模型)是从学生的答题表现序列来推断其内在掌握知识情况的隐马尔科夫模型,它在智能辅导系统广泛使用。传统的KT模型通常只适用于对单个知识点能力的建模,且不能对题目难度和分辨度进行推断。但在现今智慧教育环境中,题目序列往往包含多个知识点,需要利用这些知识点对学生能力变化进行综合的评估和分析。为此,文章使用Logistic回归方法扩展KT模型的参数,提出三个面向多知识点的KT模型的改进模型:KTLR-GS模型、KTLR-LFID模型和KTLR-FP模型,分别把猜对概率、犯错概率、题目难度、观测参数和状态转移参数等作为特征引入,从而实现了在多知识点背景下对学生学习能力和综合能力的准确评价,超过了传统KT模型的分析性能。同时,文章还开发了基于xAPI的学习行为分析与可视化系统,集学习行为采集、存储、分析和可视化为一体,利用面向多知识点的KT模型分析学生答题序列中的知识掌握动态变化,并利用可视化的技术帮助教师对每个学生的学习情况进行及时判断。


【英文摘要】:

Knowledge Tracing Model (KT Model) is a Hidden Markov Model for inferring students' knowledge from students' answers, which is widely used in intelligent tutoring system. The traditional KT Model are usually only applicable to Modeling the capabilities of a single point of knowledge and cannot infer the difficulty and resolution of the questions. However, in today's intelligent educational system, the questions often contain multiple knowledge points, which can be used to comprehensively evaluate and analyze the changes of students' abilities. Therefore, this paper uses logistic regression method to extend the parameters of KT Model and proposes three improved models for multiple knowledge points: KTLR-GS Model, KTLR-LFID Model and KTLR-FP Model. In those models, probability of guessing correctly, probability of making mistakes, difficulty of the problems, observation parameter and state transfer parameter are all introduced as features. As a result, those models achieve the accurate evaluation of students' learning abilities and comprehensive abilities under the background of multiple knowledge points, which exceeds the performance of the traditional KT Model. Meanwhile, this paper develops a xAPI-based learning behavior analysis and visualization system, integrating the acquisition, storage, analysis and visualization of learning behavior data, which uses KT Models for multiple knowledge points to analyze the dynamic changes of students' knowledge in their answers, and help teachers to make judgments about every student on time through visualization.

网站首页 | 联系我们 | 意见建议| 杂志订阅| 管理部登录

版权所有 © 电化教育研究 CopyRight ©e-EDUCATION RESEARCH 2011-2018 电话:0931-7971823 地址:西北师范大学《电化教育研究》杂志社