Baysian Evaluation System based on Teachers 'Wisdom Using Network Model

(1)运用贝叶斯层次建模理论对网络模型进行了分析。同时，利用Gibbs采样器建立了适用于层次分析法的分析方法。为了分析学习者的内部过程，需要进行复杂的建模，但可用的数据往往是分类数据。因此，我们需要一个包含潜变量分布和先验分布的层次模型。同时，Gibbs抽样也是一个非常有用的工具来获得可靠的参数估计。(2)开发了一个贝叶斯网络模型来识别学习者的错误规则，以及更一般的学习者在学习地图中的状态。该系统计算每个可能的错误规则的后验概率。作为另一个应用，开发了以连续变量为节点的医疗网络诊断系统。该系统易于修改，便于对教育数据进行处理。(3)为了获取教师有用的信息，开发了多学科回答的项目反应模型。这种分析可以区分相同分数的能力参数，并可以为每个测试项目提供更详细的信息。

(1) The Baysian Theory of Hierarchical Modeling was developed to analyze the network model. Also, the suitable analysis method for the hierarchical method was developed using the Gibbs Sampler. In order to analyze the internal process of the leaner, the complex modeling is necessary, but the available data is often categorical data. Therefore, we need the hierarchical model which includes the distributions of latent variables and the prior distributions. Also, the Gibbs sampling is a very useful tool to obtain the reliable estimates of parameters.(2) A Baysian network model was developed to identify learners' bug rules and more generally learner's status in the learning map. This system calculates the posterior probability for each of the possible bug rules. Also, as another application, the medical network diagnosis system which includes the continuous variables as nodes was developed. This system can be modified easily to handle the educational data.(3) An Item Response Model for the polychotomous responses was developed to obtain useful information for teachers. This analysis can discriminates the ability parameters of the same score, and can provide the more detailed information for each of test items.