摘要
目前有不少的研究指出,利用贝氏网络诊断学生的错误类型(Bug)以及子技能(Skill)的学习情况,其成效良好。结合计算机进行仿真研究,更可以了解在不同模式设计下的贝氏网络之预测精准度,以提供建置计算机化测验系统之参考。因此,研究的目的为评估不同模式之贝氏网络的诊断成效。研究结果显示,将选择题选项对应之错误类型视为试题当作证据时,比单纯选择题时的预测精准度提升5.2%,而加入专家知识结构能够再提升0.8%。因此,在建置贝氏网络为基础的微积分计算机化测验时,应参考模式四之设计。
Bayesian network can predict and diagnosis efficiently. Many researches have showed that results of learning with Bayesian network are good. Combining the simulation study with computers, it is much easier to understand the precision of the prediction for dif- ferent Bayesian network models. Therefore, the purpose of the research is to evaluate the diagnosis effectiveness between the different models of Bayesian network. The research shows, that by using the Bugs which correlates to the multiple choice questions as new inputs, the precision of the prediction is raised 5.2% comparing to the precision of the original questions,and it raised 0.8% more when inclu- ding knowledge structure in the precision of the original questions. Therefore, to construct an calculus computerized test based on Bayes- ian network, model 4 should be referred.
出处
《心理学探新》
CSSCI
2012年第5期414-416,共3页
Psychological Exploration
关键词
贝氏网络
计算机化测验
认知诊断
微分基本公式
Bayesian network
computerized test
cognitive diagnosis model
differential basic formula
