基于K2算法的属性层级结构学习研究

诊断测验所考察的属性之间往往存在某种层级关系,然而基于专家经验获得的属性层级关系易出现分歧或错误.该文将属性掌握模式作为输入,考察K2算法在不同阈值条件下学习得到属性层级结构的准确性.模拟研究和实证数据分析的结果表明:K2算法对属性层级结构的学习有较高的成功率,并且K2算法对于4种基本层级结构有不同的敏感性,其中线性型和发散型对阈值的敏感性较低,而收敛型和无结构型对于阈值的敏感性较高.

The Attribute Hierarchical Structure Learning Based on K2 Algorithm

There is often some hierarchical relationship among the attributes measured by the diagnostic test,but attribute hierarchical relationships are often difficult to obtain.Experience-based attribute-relationships are prone to errors.The attribute mastering model is taken as input,and the accuracy of the attribute hierarchical structure learned by the K2 algorithm under different conditions is evaluated.Simulation research and empirical data analysis results show that the K2 algorithm has a high accuracy rate for the learning of the attribute hierarchy,and the K2 algorithm has different sensitivities to the four basic hierarchical structures,among which the linear and divergent types are sensitive to the threshold,while the convergence type and unstructured type have higher sensitivity to the threshold.